1 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
3 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
6 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
8 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
9 early gimple folding of vec_splat().
10 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
11 * gimple-fold.h: Add an extern define for tree_vec_extract().
13 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
15 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
16 wrappers around TREE_TYPE comparisons.
18 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
21 * config/s390/predicates.md: Add nonsym_memory_operand.
22 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
23 contains a SYMBOL_REF, load it into an intermediate pseudo.
24 (s390_emit_compare_and_swap): Legitimize operand.
25 * config/s390/s390.md: Use the new nonsym_memory_operand
26 with UNSPECV_CAS patterns.
28 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
31 * config/s390/s390-passes.def: New file.
32 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
34 (make_pass_s390_early_mach): Add declaration.
35 * config/s390/s390.c (make_pass_s390_early_mach):
36 (s390_option_override): Remove dynamic registration.
37 * config/s390/t-s390: Add s390-passes.def.
39 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
41 * config/s390/s390.c (s390_decompose_constant_pool_ref):
42 Remove UNSPEC_LTREL_BASE check.
43 (annotate_constant_pool_refs): Likewise.
44 (find_constant_pool_ref): Likewise.
45 (find_ltrel_base): Removed.
46 (replace_ltrel_base): Removed.
47 (s390_mainpool_finish): Remove replace_ltrel_base call.
48 (s390_chunkify_start): Remove pending LTREL_BASE logic.
49 (s390_chunkify_finish): Remove replace_ltrel_base call.
50 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
52 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
55 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
56 to speculation_safe_value_not_needed.
58 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
59 Bernd Schmidt <bernds_cb1@t-online.de>
61 * config/nvptx/nvptx-opts.h: New file.
62 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
63 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
65 (TARGET_SM35): New macro.
66 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
68 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
71 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
73 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
75 * config/i386/i386.md (truncdfsf2): Remove expander.
76 (truncdfsf2_with_temp): Ditto.
77 (truncxf<mode>2): Ditto.
78 (*truncdfsf_fast_mixed): Remove insn pattern.
79 (*truncdfsf_fast_i387): Ditto.
80 (*truncdfsf_mixed): Ditto.
81 (*truncdfsf_i387): Ditto.
82 (*truncdfsf2_i387_1): Ditto.
83 (*truncxfsf2_mixed): Ditto.
84 (*truncxfdf2_mixed): Ditto.
85 (*truncxf<mode>2_i387_noop): Ditto. Update callers
86 to call gen_truncxf<mode>2 instead.
87 (*truncxf<mode>2_i387): Remove.
88 (reg->reg splitters): Remove splitter pattern.
89 (reg->mem splitters): Ditto.
91 (truncdfsf2): New insn pattern.
92 (truncxf<mode>2): Ditto.
94 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
96 * tree-ssa-math-opts.c (is_mult_by): New function.
97 (is_square_of): Use the above.
98 (optimize_recip_sqrt): New function.
99 (pass_cse_reciprocals::execute): Use the above.
101 2018-09-05 Richard Biener <rguenther@suse.de>
104 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
105 to zero-init the emplaced vec.
107 2018-09-05 Martin Liska <mliska@suse.cz>
109 PR tree-optimization/87205
110 * tree-switch-conversion.c (pass_lower_switch::execute):
111 Group cases for switch statements.
113 2018-09-05 Richard Biener <rguenther@suse.de>
115 PR tree-optimization/87217
116 * tree-ssa-sccvn.c (vuse_valueize): New.
117 (vn_reference_lookup_pieces): Use it.
118 (vn_reference_lookup): Likewise.
120 2018-09-05 Nathan Sidwell <nathan@acm.org>
123 * stor-layout.c (place_field): Scan forwards to check last
124 bitfield when ms_bitfield_placement is in effect.
126 2018-09-05 Richard Biener <rguenther@suse.de>
129 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
132 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
133 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
135 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
136 * config.gcc (extra_objs): Build it.
137 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
139 * config/aarch64/aarch64-passes.def
140 (pass_tag_collision_avoidance): New pass.
141 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
142 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
143 (aarch64_classify_address): Remove static qualifier.
144 (aarch64_address_info, aarch64_address_type): Move to...
145 * config/aarch64/aarch64-protos.h: ... here.
146 (make_pass_tag_collision_avoidance): New function.
147 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
150 2018-09-05 Martin Liska <mliska@suse.cz>
152 * doc/gcov.texi: Update documentation of humar
154 * gcov.c (format_count): Print one decimal place, it provides
155 more fine number of situations like '1G' vs. '1.4G'.
157 2018-09-05 Martin Liska <mliska@suse.cz>
160 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
161 * optc-gen.awk: Allow 'Var' for Deprecated options in order
162 to generate a MASK value.
164 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
167 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
168 if frame pointer isn't used.
169 (compute_frame_pointer_to_fb_displacement): Likewise.
171 2018-09-04 Jakub Jelinek <jakub@redhat.com>
174 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
175 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
176 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
177 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
178 and OPTION_MASK_ISA_XSAVEC_UNSET.
180 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
182 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
183 XOR operations in NAND case.
185 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
187 * wide-int-range.cc (wide_int_range_convert): New.
188 * wide-int-range.h (wide_int_range_convert): New.
189 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
190 code into wide_int_range_convert.
191 (extract_range_into_wide_ints): Do not munge anti range constants
192 into the entire domain. Just return the range back.
194 2018-09-04 Martin Liska <mliska@suse.cz>
196 * genmatch.c (output_line_directive): Add new argument
198 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
200 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
202 * doc/invoke.texi (Option Summary): Add whitespace.
204 * doc/invoke.texi (Option Summary): Add -Waligned-new.
206 2018-09-04 Richard Biener <rguenther@suse.de>
208 PR tree-optimization/87211
209 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
210 backedge value we're supposed to treat as VARYING also number
211 the PHI to VARYING in case it got a different value-number already.
213 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
215 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
216 (extract_range_from_binary_expr_1): Do not call
217 vrp_can_optimize_bit_op.
218 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
220 (wide_int_range_get_mask_and_bounds): New.
221 (wide_int_range_optimize_bit_op): New.
222 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
223 (wide_int_range_bit_and): Same.
224 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
225 (wide_int_range_optimize_bit_op): New.
226 (wide_int_range_get_mask_and_bounds): New.
228 2018-09-04 Richard Biener <rguenther@suse.de>
230 PR tree-optimization/87176
231 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
232 variable. When value-numbering a virtual PHI node make sure
233 to not value-number to the backedge value.
235 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
237 * doc/extend.texi (Long Long, Hex Floats): Document support for
238 long long and hex floats in more recent versions of ISO C++.
240 2018-09-03 Richard Biener <rguenther@suse.de>
242 PR tree-optimization/87177
243 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
246 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
248 * bb-reorder.c (edge_order): Convert to C-qsort-style
249 tri-state comparator.
250 (reorder_basic_blocks_simple): Change std::stable_sort to gcc_stablesort.
252 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
254 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
255 tri-state comparator.
256 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
258 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
260 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
261 (mergesort): ... here as maximum count for using netsort.
262 (gcc_qsort): Set nlim to 3 if stable sort is requested.
263 (gcc_stablesort): New.
264 * system.h (gcc_stablesort): Declare.
266 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
268 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
269 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
270 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
272 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
274 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
275 lxsdx and stxsdx alternatives.
276 (*mov<mode>_hardfloat64): Ditto.
277 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
279 2018-09-03 Richard Biener <rguenther@suse.de>
281 PR tree-optimization/87200
282 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
285 2018-09-03 Martin Liska <mliska@suse.cz>
287 PR tree-optimization/87201
288 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
289 Fix parenthesis in an expression.
291 2018-09-03 Richard Biener <rguenther@suse.de>
293 PR tree-optimization/87197
294 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
295 visited. CSE the VN_INFO hashtable lookup.
297 PR tree-optimization/87169
298 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
299 iterating make sure there's no extra backedges from irreducible
300 regions feeding the header. Mark the destination block
303 2018-09-03 Martin Liska <mliska@suse.cz>
306 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
307 * common/common-targhooks.c (default_get_valid_option_values):
309 * common/common-targhooks.h (default_get_valid_option_values):
311 * common/config/i386/i386-common.c: Move processor_target_table
313 (ix86_get_valid_option_values): New function.
314 (TARGET_GET_VALID_OPTION_VALUES): New macro.
315 * config/i386/i386.c (struct ptt): Move to i386-common.c.
316 (PTA_*): Move all defined masks into i386-common.c.
317 (ix86_function_specific_restore): Use new processor_cost_table.
318 * config/i386/i386.h (struct ptt): Moved from i386.c.
319 (struct pta): Likewise.
320 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
321 * doc/tm.texi.in: Likewise.
322 * opt-suggestions.c (option_proposer::suggest_option):
323 Pass prefix to build_option_suggestions.
324 (option_proposer::get_completions): Likewise.
325 (option_proposer::build_option_suggestions): Use the new target
327 * opts.c (struct option_help_tuple): New struct.
328 (print_filtered_help): Use the new target hook.
330 2018-09-03 Martin Liska <mliska@suse.cz>
333 * predict.c (set_even_probabilities): Add likely_edges
334 argument and handle cases where we have precisely one
336 (combine_predictions_for_bb): Catch also likely_edges.
337 (tree_predict_by_opcode): Handle gswitch statements.
338 * tree-cfg.h (find_case_label_for_value): New declaration.
339 (find_taken_edge_switch_expr): Likewise.
340 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
341 Find pivot in decision tree based on probabily, not by number of
344 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
346 * doc/standards.texi (Standards): Update Objective-C reference.
348 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
350 * doc/install.texi (Prerequisites): Update link for MPC.
352 2018-09-01 Michael Matz <matz@suse.de>
354 PR tree-optimization/87074
355 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
356 PHIs for outer-loop uses.
358 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
360 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
361 * doc/invoke.texi (C Dialect Options): Ditto.
363 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
365 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
367 2018-08-31 Richard Biener <rguenther@suse.de>
369 PR tree-optimization/87168
370 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
371 (rpo_elim::eliminate_avail): When OP was not visited it must
374 2018-08-31 David Malcolm <dmalcolm@redhat.com>
376 * tree-vrp.c (copy_value_range): Convert param "from" from
377 "value_range *" to "const value_range *".
378 (range_is_null): Likewise for param "vr".
379 (range_int_cst_p): Likewise.
380 (range_int_cst_singleton_p): Likewise.
381 (symbolic_range_p): Likewise.
382 (value_ranges_intersect_p): Likewise for both params.
383 (value_range_nonnegative_p): Likewise for param "vr".
384 (value_range_constant_singleton): Likewise.
385 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
386 (extract_range_into_wide_ints): Likewise for param "vr".
387 (extract_range_from_multiplicative_op): Likewise for params "vr0"
389 (vrp_can_optimize_bit_op): Likewise.
390 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
392 (extract_range_from_unary_expr): Likewise.
393 (debug_value_range): Likewise for param "vr".
394 (value_range::dump): Add "const" qualifier.
395 (vrp_prop::check_array_ref): Convert local "vr" from
396 "value_range *" to "const value_range *".
397 (vrp_prop::check_mem_ref): Likewise.
398 (vrp_prop::visit_stmt): Likewise for local "old_vr".
399 (vrp_intersect_ranges_1): Likewise for param "vr_1".
400 (vrp_intersect_ranges): Likewise.
401 (simplify_stmt_for_jump_threading): Likewise for local "vr".
402 (vrp_prop::vrp_finalize): Likewise.
403 * tree-vrp.h (value_range::dump): Add "const" qualifier.
404 (vrp_intersect_ranges): Add "const" qualifier to params as above.
405 (extract_range_from_unary_expr): Likewise.
406 (value_range_constant_singleton): Likewise.
407 (symbolic_range_p): Likewise.
408 (copy_value_range): Likewise.
409 (extract_range_from_binary_expr_1): Likewise.
410 (range_int_cst_p): Likewise.
411 (vrp_set_zero_nonzero_bits): Likewise.
412 (range_int_cst_singleton_p): Likewise.
414 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
416 * config/aarch64/arm_neon.h (vabsd_s64): New.
417 (vnegd_s64): Likewise.
419 2018-08-31 Martin Jambor <mjambor@suse.cz>
421 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
423 2018-08-31 Martin Liska <mliska@suse.cz>
425 * ipa-icf.c (sem_item::add_type): Use
426 sem_item::m_type_hash_cache.
427 * ipa-icf.h: Move the cache from sem_item_optimizer
430 2018-08-31 Nathan Sidwell <nathan@acm.org>
432 * doc/extend.texi (Backwards Compatibility): Remove implicit
433 extern C leeway of () being (...).
435 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
437 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
439 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
443 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
445 2018-08-31 Jakub Jelinek <jakub@redhat.com>
448 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
449 gen_int_mode. Formatting fixes.
451 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
453 * target.def (custom_function_descriptors): Improve documentation.
454 * doc/tm.texi.in (Trampolines): Expand discussion of function
455 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
456 beginning of the section.
457 * doc/tm.texi: Regenerated.
459 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
461 * cfg.h (class auto_edge_flag): Spell out the template-id of the
462 base class in the initializer list. This is a workaround for
463 building with older GCC.
464 (class auto_bb_flag): Likewise.
466 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
468 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
469 (altivec_vcmpequ<VI_char>_p): Remove star.
470 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
472 (expand_strncmp_vec_sequence): New function.
473 (emit_final_str_compare_vec): New function.
474 (expand_strn_compare): Add support for vector strncmp.
475 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
476 length specification to bytes.
477 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
478 (vcmpnezb_p): New pattern.
479 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
480 for option -mstring-compare-inline-limit.
482 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
484 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
485 (PTA_SKYLAKE): Add PTA_AES.
486 (PTA_GOLDMONT): Likewise.
488 2018-08-29 Jan Hubicka <jh@suse.cz>
491 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
492 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
494 2018-08-29 Jan Hubicka <jh@suse.cz>
496 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
498 (hash_tree): Do not visit TYPE_STUB_DECL.
499 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
500 stream TYPE_STUB_DECL.
501 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
502 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
503 after free_lang_data.
504 (type_in_anonymous_namespace_p): Likewise.
506 2018-08-29 Jan Hubicka <jh@suse.cz>
508 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
509 comment that it has to be even number.
510 (class sreal): Change m_sig type to int32_t.
511 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
512 int64_t for temporary calculations.
513 (sreal_verify_basics): Drop one bit from minimum and maximum.
515 2018-08-30 Richard Biener <rguenther@suse.de>
517 PR tree-optimization/87147
518 * tree-ssa-sccvn.c (SSA_VISITED): New function.
519 (visit_phi): When the degenerate result is from the backedge and
520 we didn't visit its definition yet drop to VARYING.
521 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
523 2018-08-29 Jan Hubicka <jh@suse.cz>
525 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
527 (hash_tree): Likewise.
529 2018-08-29 Jan Hubicka <jh@suse.cz>
531 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
532 and TYPE_NEXT_REF_TO.
534 2018-08-29 Jan Hubicka <jh@suse.cz>
536 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
537 comment that it has to be even number.
538 (class sreal): Change m_sig type to int32_t.
539 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
540 int64_t for temporary calculations.
541 (sreal_verify_basics): Drop one bit from minimum and maximum.
543 2018-08-30 Tamar Christina <tamar.christina@arm.com>
545 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
547 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
550 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
551 if to_add is negative.
553 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
556 * builtins.c (c_strlen): Improve range checks.
558 2018-08-29 Martin Sebor <msebor@redhat.com>
559 Jeff Law <law@redhat.com>
561 PR tree-optimization/86714
562 PR tree-optimization/86711
563 * builtins.c (c_strlen): Add arguments to call to string_constant.
564 * expr.c (string_constant): Add argument. Detect missing nul
565 terminator and outermost declaration it's missing in.
566 * expr.h (string_constant): Add argument.
567 * fold-const.c (read_from_constant_string): Add arguments to call to
569 (c_getstr): Likewise.
570 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
572 * tree-ssa-strlen.c (get_stridx): Likewise.
574 2018-08-29 Jan Hubicka <jh@suse.cz>
576 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
577 Do not stream DECL_VINDEX.
578 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
579 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
580 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
582 2018-08-29 Richard Biener <rguenther@suse.de>
584 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
585 virtual operands that are not default defs to honor region
587 (rpo_vn_valueize): Remove ineffective code here.
589 2018-08-29 Richard Biener <rguenther@suse.de>
591 PR tree-optimization/87132
592 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
593 when skipping defs reachable over backedges.
595 2018-08-29 Richard Biener <rguenther@suse.de>
597 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
598 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
599 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
600 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
601 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
602 (vn_reference_lookup_3): Remove use of const_parms.
603 (free_rpo_vn): Do not free const_parms.
604 (do_rpo_vn): Do not call init_const_parms.
605 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
606 SSA_NAME_POINTS_TO_READONLY_MEMORY.
607 (call_may_clobber_ref_p_1): Likewise.
609 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
612 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
614 (-ftree-scev-cprop): Document.
616 2018-08-29 Jan Hubicka <jh@suse.cz>
618 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
620 (sreal constructor): Update.
621 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
622 sreal:operator/): Update.
624 2018-08-29 Martin Liska <mliska@suse.cz>
626 * tree-switch-conversion.c (switch_conversion::expand):
627 Strenghten assumption about gswitch statements.
629 2018-08-29 Richard Biener <rguenther@suse.de>
631 PR tree-optimization/87117
632 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
633 re-value-number released SSA VDEFs.
635 2018-08-29 Richard Biener <rguenther@suse.de>
637 PR tree-optimization/87126
638 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
640 2018-08-28 Jim Wilson <jimw@sifive.com>
642 * config/riscv/pic.md: Rewrite.
643 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
645 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
646 (SOFTF, default_load, softload, softstore): New.
648 2018-08-28 Jeff Law <law@redhat.com>
650 * fold-const.c (fold_binary_loc): Remove recently added assert.
652 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
654 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
655 to OP parmeter of generated function.
657 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
659 PR tree-optimization/87009
660 * match.pd: Add boolean optimizations.
662 2018-08-28 Martin Sebor <msebor@redhat.com>
665 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
666 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
667 (pass_walloca::gate): Use it.
668 (alloca_call_type): Same.
669 (pass_walloca::execute): Same.
670 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
672 2018-08-28 David Malcolm <dmalcolm@redhat.com>
674 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
675 GCC_VERSION for usage of "__gcc_dump_printf__" format from
678 2018-08-28 Richard Biener <rguenther@suse.de>
680 PR tree-optimization/87124
681 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
682 constants before looking up avail.
684 2018-08-28 Jakub Jelinek <jakub@redhat.com>
687 * calls.c (maybe_warn_nonstring_arg): Punt early if
688 warn_stringop_overflow is zero. Don't call get_range_strlen
689 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
690 Swap comparison operands to have constants on rhs. Only use
691 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
694 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
696 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
697 use of tree_to_shwi. Remove duplicated test for the size being
698 a whole number of bytes.
700 2018-08-28 Richard Biener <rguenther@suse.de>
702 PR tree-optimization/87117
703 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
704 Handle removed stmt without LHS (GIMPLE_NOP).
706 2018-08-28 Richard Biener <rguenther@suse.de>
708 PR tree-optimization/87117
709 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
710 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
712 2018-08-28 Richard Biener <rguenther@suse.de>
714 PR tree-optimization/87117
715 * tree-ssa-pre.c (compute_avail): Do not make expressions
716 with predicated values available.
717 (get_expr_value_id): Assert we do not run into predicated value
720 2018-08-28 Richard Biener <rguenther@suse.de>
722 PR tree-optimization/87117
723 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
724 get virtual operands.
725 (get_expr_operands): Handle STRING_CST like other decls.
727 2018-08-28 Martin Liska <mliska@suse.cz>
729 * tree.h: Update documentation of fndecl_built_in_p
733 2018-08-27 Jeff Law <law@redhat.com>
734 PR tree-optimization/87110
735 * tree-ssa-dse.c (compute_trims): Handle non-constant
738 2018-08-27 Martin Sebor <msebor@redhat.com>
740 PR tree-optimization/86914
741 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
743 2018-08-27 Martin Sebor <msebor@redhat.com>
745 PR tree-optimization/87112
746 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
747 the type of the bound argument.
749 2018-08-27 Jeff Law <law@redhat.com>
751 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
752 type does not have a TYPE_SIZE_UNIT.
754 2018-08-27 Steve Ellcey <sellcey@cavium.com>
756 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
757 with include of backend.h.
759 2018-08-27 Richard Biener <rguenther@suse.de>
761 PR tree-optimization/86927
762 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
763 use const cond reduction code.
765 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
767 PR tree-optimization/85758
768 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
770 2018-08-27 David Malcolm <dmalcolm@redhat.com>
773 * diagnostic-show-locus.c (class layout_range): Update for
774 conversion of show_caret_p to a tri-state.
775 (layout_range::layout_range): Likewise.
776 (make_range): Likewise.
777 (layout::maybe_add_location_range): Likewise.
778 (layout::should_print_annotation_line_p): Don't show annotation
779 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
780 (layout::get_state_at_point): Update for conversion of
781 show_caret_p to a tri-state. Bail out early for
782 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
783 underlining or source colorization.
784 (gcc_rich_location::add_location_if_nearby): Update for conversion
785 of show_caret_p to a tri-state.
786 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
787 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
789 (selftest::test_one_liner_labels): Likewise.
790 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
791 conversion of show_caret_p to a tri-state.
792 * pretty-print.c (text_info::set_location): Likewise.
793 * pretty-print.h (text_info::set_location): Likewise.
794 * substring-locations.c (format_warning_n_va): Likewise.
795 * tree-diagnostic.c (default_tree_printer): Likewise.
796 * tree-pretty-print.c (newline_and_indent): Likewise.
798 2018-08-27 David Malcolm <dmalcolm@redhat.com>
801 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
802 line above for line-insertion fix-it hints.
803 (selftest::test_fixit_insert_containing_newline): Update the
804 expected results, and add a test with line-numbering enabled.
806 2018-08-27 Martin Liska <mliska@suse.cz>
809 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
810 params with DECL_HAS_VALUE_EXPR_P.
812 2018-08-27 Martin Liska <mliska@suse.cz>
814 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
815 selected expansion strategy.
817 2018-08-27 Martin Liska <mliska@suse.cz>
819 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
820 * builtins.c (is_builtin_fn): Likewise.
821 * attribs.c (diag_attr_exclusions): Use new function
822 fndecl_built_in_p and remove check for FUNCTION_DECL if
824 (builtin_mathfn_code): Likewise.
825 (fold_builtin_expect): Likewise.
826 (fold_call_expr): Likewise.
827 (fold_builtin_call_array): Likewise.
828 (fold_call_stmt): Likewise.
829 (set_builtin_user_assembler_name): Likewise.
830 (is_simple_builtin): Likewise.
831 * calls.c (gimple_alloca_call_p): Likewise.
832 (maybe_warn_nonstring_arg): Likewise.
833 * cfgexpand.c (expand_call_stmt): Likewise.
834 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
835 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
836 (cgraph_node::verify_node): Likewise.
837 * cgraphclones.c (build_function_decl_skip_args): Likewise.
838 (cgraph_node::create_clone): Likewise.
839 * config/arm/arm.c (arm_insert_attributes): Likewise.
840 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
841 * dse.c (scan_insn): Likewise.
842 * expr.c (expand_expr_real_1): Likewise.
843 * fold-const.c (operand_equal_p): Likewise.
844 (fold_binary_loc): Likewise.
845 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
846 * gimple-low.c (lower_stmt): Likewise.
847 * gimple-pretty-print.c (dump_gimple_call): Likewise.
848 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Likewise.
849 * gimple.c (gimple_build_call_from_tree): Likewise.
850 (gimple_call_builtin_p): Likewise.
851 (gimple_call_combined_fn): Likewise.
852 * gimplify.c (gimplify_call_expr): Likewise.
853 (gimple_boolify): Likewise.
854 (gimplify_modify_expr): Likewise.
855 (gimplify_addr_expr): Likewise.
856 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
857 * ipa-cp.c (determine_versionability): Likewise.
858 * ipa-fnsummary.c (compute_fn_summary): Likewise.
859 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
860 * ipa-split.c (visit_bb): Likewise.
861 (split_function): Likewise.
862 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
863 * lto-cgraph.c (input_node): Likewise.
864 * lto-streamer-out.c (write_symbol): Likewise.
865 * omp-low.c (setjmp_or_longjmp_p): Likewise.
866 (lower_omp_1): Likewise.
867 * predict.c (strip_predict_hints): Likewise.
868 * print-tree.c (print_node): Likewise.
869 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
870 * trans-mem.c (is_tm_irrevocable): Likewise.
871 (is_tm_load): Likewise.
872 (is_tm_simple_load): Likewise.
873 (is_tm_store): Likewise.
874 (is_tm_simple_store): Likewise.
875 (is_tm_abort): Likewise.
876 (tm_region_init_1): Likewise.
877 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
878 * tree-cfg.c (verify_gimple_call): Likewise.
879 (move_stmt_r): Likewise.
880 (stmt_can_terminate_bb_p): Likewise.
881 * tree-eh.c (lower_eh_constructs_2): Likewise.
882 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
883 * tree-inline.c (remap_gimple_stmt): Likewise.
885 (estimate_num_insns): Likewise.
886 (fold_marked_statements): Likewise.
887 * tree-sra.c (scan_function): Likewise.
888 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
889 (optimize_stack_restore): Likewise.
890 (pass_fold_builtins::execute): Likewise.
891 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
892 (mark_all_reaching_defs_necessary_1): Likewise.
893 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
894 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
895 (pass_forwprop::execute): Likewise.
896 * tree-ssa-loop-im.c (stmt_cost): Likewise.
897 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
898 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
899 * tree-ssa-strlen.c (get_string_length): Likewise.
900 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
901 (find_func_aliases_for_call): Likewise.
902 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
903 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
904 * tree-tailcall.c (find_tail_calls): Likewise.
905 * tree.c (need_assembler_name_p): Likewise.
906 (free_lang_data_in_decl): Likewise.
907 (get_call_combined_fn): Likewise.
908 * ubsan.c (is_ubsan_builtin_p): Likewise.
909 * varasm.c (incorporeal_function_p): Likewise.
910 * tree.h (DECL_BUILT_IN): Remove and replace with
912 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
913 (fndecl_built_in_p): New.
915 2018-08-27 Martin Liska <mliska@suse.cz>
917 PR tree-optimization/86847
918 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
919 Dump also subtree probability.
920 (switch_decision_tree::do_jump_if_equal): New function.
921 (switch_decision_tree::emit_case_nodes): Handle special
922 situations in balanced tree that can be emitted much simpler.
923 Fix calculation of probabilities that happen in tree expansion.
924 * tree-switch-conversion.h (struct cluster): Add
926 (struct simple_cluster): Likewise.
927 (struct case_tree_node): Add new function has_child.
928 (do_jump_if_equal): New.
930 2018-08-27 Martin Liska <mliska@suse.cz>
932 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
933 Add new argument to bit_test_cluster constructor.
934 (bit_test_cluster::emit): Set bits really number of values
936 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
937 probability argument.
938 * tree-switch-conversion.h (struct bit_test_cluster):
939 Add m_handles_entire_switch.
941 2018-08-27 Martin Liska <mliska@suse.cz>
943 PR tree-optimization/86702
944 * tree-switch-conversion.c (jump_table_cluster::emit):
945 Make probabilities even for values in jump table
946 according to number of cases handled.
947 (switch_decision_tree::compute_cases_per_edge): Pass
948 argument to reset_out_edges_aux function.
949 (switch_decision_tree::analyze_switch_statement): Likewise.
950 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
953 2018-08-27 Martin Liska <mliska@suse.cz>
955 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
956 cfun argument explicitly.
957 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
958 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
959 function gimple_switch_default_bb.
960 (convert_switch_statements):
962 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
963 * stmt.c (label_to_block_fn): Use label_to_block and pass
964 cfun argument explicitly and use gimple_switch_label_bb.
965 (expand_case): Likewise.
966 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
967 cfun argument explicitly. Likewise.
968 (make_edges_bb): Likewise.
969 (make_cond_expr_edges): Likewise.
970 (get_cases_for_edge): Likewise.
971 (make_gimple_switch_edges): Likewise.
972 (label_to_block_fn): Likewise.
973 (label_to_block): Likewise.
974 (make_goto_expr_edges): Likewise.
975 (make_gimple_asm_edges): Likewise.
976 (main_block_label): Likewise.
977 (group_case_labels_stmt): Likewise.
978 (find_taken_edge_computed_goto): Likewise.
979 (find_taken_edge_switch_expr): Likewise.
980 (gimple_verify_flow_info): Likewise.
981 (gimple_redirect_edge_and_branch): Likewise.
982 (gimple_switch_label_bb): New function.
983 (gimple_switch_default_bb): Likewise.
984 (gimple_switch_edge): Likewise.
985 (gimple_switch_default_edge): Likewise.
986 * tree-cfg.h (label_to_block_fn): Remove and replace ...
987 (label_to_block): ... with this.
988 (gimple_switch_label_bb): New.
989 (gimple_switch_default_bb): Likewise.
990 (gimple_switch_edge): Likewise.
991 (gimple_switch_default_edge): Likewise.
992 * tree-cfgcleanup.c (convert_single_case_switch): Use
993 new gimple functions and pass new argument to label_to_block.
994 (cleanup_control_flow_bb):
995 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
996 cfun argument explicitly.
997 (make_eh_edges): Likewise.
998 (redirect_eh_dispatch_edge): Likewise.
999 (lower_resx): Likewise.
1000 (lower_eh_dispatch): Likewise.
1001 (maybe_remove_unreachable_handlers): Likewise.
1002 (unsplit_eh): Likewise.
1003 (cleanup_empty_eh): Likewise.
1004 (verify_eh_edges): Likewise.
1005 (verify_eh_dispatch_edge): Likewise.
1006 * tree-ssa-dom.c (record_edge_info): Likewise.
1007 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
1008 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
1009 (thread_through_normal_block): Likewise.
1010 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
1011 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
1012 * tree-switch-conversion.c (switch_conversion::collect): Use new
1014 (switch_conversion::check_final_bb): Likewise.
1015 (switch_conversion::gather_default_values): Pass new argument
1017 (switch_conversion::build_constructors): Likewise.
1018 (switch_decision_tree::compute_cases_per_edge): Use new
1019 gimple_switch_edge function.
1020 (switch_decision_tree::analyze_switch_statement): Pass new argument
1022 (switch_decision_tree::try_switch_expansion): Use
1023 gimple_switch_default_edge.
1024 * tree-vrp.c (find_switch_asserts): Pass new argument
1026 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
1027 (vr_values::simplify_switch_using_ranges): Likewise.
1029 2018-08-27 Richard Biener <rguenther@suse.de>
1031 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
1032 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
1034 * tree-ssa-sccvn.h (struct vn_pval): New structure.
1035 (struct vn_nary_op_s): Add unwind_to member. Add
1036 predicated_values flag and put result into a union together
1037 with a linked list of vn_pval.
1038 (struct vn_ssa_aux): Add name member to make maintaining
1039 a map of SSA name to vn_ssa_aux possible. Remove no longer
1040 needed info, dfsnum, low, visited, on_sccstack, use_processed
1041 and range_info_anti_range_p members.
1042 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
1043 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
1045 (vn_valueize): New global.
1046 (vn_context_bb): Likewise.
1047 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
1048 VN_INFO_PTR_INFO): Remove.
1049 * tree-ssa-sccvn.c: ... (rewrite)
1050 (pass_fre::execute): For -O2+ initialize loops and run
1051 RPO VN in optimistic mode (iterating). For -O1 and -Og
1052 run RPO VN in non-optimistic mode.
1053 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
1054 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
1055 * doc/invoke.texi (sccvn-max-scc-size): Remove.
1056 (rpo-vn-max-loop-depth): Document.
1057 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
1058 when valuezing the VUSE signals we walked out of the region.
1059 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
1060 (phi_translate): Set VN context block to use for availability
1062 (compute_avail): Likewise.
1063 (pre_valueize): New function.
1064 (pass_pre::execute): Adjust to the RPO VN API.
1066 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
1067 (propagate_constants_for_unrolling): Remove.
1068 (tree_unroll_loops_completely): Perform value-numbering
1069 on the unrolled bodies loop parent.
1071 2018-08-27 Richard Biener <rguenther@suse.de>
1073 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
1074 for partial antic compute.
1076 2018-08-27 Jakub Jelinek <jakub@redhat.com>
1078 PR rtl-optimization/87065
1079 * combine.c (simplify_if_then_else): Formatting fix.
1080 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
1082 (known_cond): Don't return const_true_rtx for vector modes. Use
1083 CONST0_RTX instead of const0_rtx. Formatting fixes.
1085 2018-08-27 Martin Liska <mliska@suse.cz>
1087 PR gcov-profile/87069
1088 * gcov.c (process_file): Record files already processed
1089 and warn about a file being processed multiple times.
1091 2018-08-27 Martin Liska <mliska@suse.cz>
1094 * config/aarch64/aarch64.c (aarch64_override_options_internal):
1095 Set default values for x_aarch64_*_string strings.
1096 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
1097 prefix. For -mabi do not print '=ABI' in help and use
1098 <option_value> format for -msve-vector-bits and -moverride
1101 2018-08-26 Jeff Law <law@redhat.com>
1103 * config/mips/frame-header-opt.c: Include "backend.h" rather than
1106 2018-08-26 Marek Polacek <polacek@redhat.com>
1108 PR c++/87029, Implement -Wredundant-move.
1109 * doc/invoke.texi: Document -Wredundant-move.
1111 2018-08-25 Martin Sebor <msebor@redhat.com>
1113 PR tree-optimization/87059
1114 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
1115 to the same type as the other.
1116 * fold-const.c (fold_binary_loc): Assert expectation.
1118 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
1120 * config/darwin.c (machopic_legitimize_pic_address): Clean up
1121 extraneous parentheses, dead code section and formatting.
1123 2018-08-24 David Malcolm <dmalcolm@redhat.com>
1126 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
1127 wide enough for jumps in the line-numbering to be visible.
1128 (layout::print_gap_in_line_numbering): New member function.
1129 (layout::calculate_line_spans): When using line numbering, merge
1130 line spans that are only 1 line apart.
1131 (diagnostic_show_locus): When printing line numbers, show gaps in
1132 line numbering directly, rather than printing headers.
1133 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
1134 line-numbering with multiple line spans.
1135 (selftest::test_fixit_insert_containing_newline_2): Add test of
1136 line-numbering, in which the spans are close enough to be merged.
1138 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
1140 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
1141 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
1142 * tree-vrp.c (range_is_nonnull): Remove.
1143 (range_includes_zero_p): Accept value_range instead of min/max.
1144 (extract_range_from_binary_expr_1): Do not early bail on
1146 Use range_includes_zero_p instead of range_is_nonnull.
1147 (extract_range_from_unary_expr): Use range_includes_zero_p instead
1148 of range_is_nonnull.
1149 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
1150 special case VR_ANTI_RANGE.
1151 (vrp_finalize): Same.
1152 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
1154 (range_is_nonnull): Remove.
1155 * vr-values.c (vrp_stmt_computes_nonzero): Use
1156 range_includes_zero_p instead of range_is_nonnull.
1157 (extract_range_basic): Pass value_range to range_includes_zero_p
1158 instead of range_is_nonnull.
1160 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
1162 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
1163 * emit-rtl.h (rtl_data): Remove return_bnd.
1164 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
1165 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
1166 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
1167 (POINTER_BOUNDS_MODE): Remove definition.
1168 (make_pointer_bounds_mode): Remove.
1169 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
1170 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
1171 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
1172 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
1173 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
1174 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
1175 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
1177 * config/i386/i386-modes.def (BND32, BND64): Remove.
1178 * config/i386/i386.c (dbx_register_map): Remove bound registers.
1179 (dbx64_register_map): Ditto.
1180 (svr4_dbx_register_map): Ditto.
1181 (indirect_thunk_bnd_needed): Remove.
1182 (indirect_thunks_bnd_used): Ditto.
1183 (indirect_return_bnd_needed): Ditto.
1184 (indirect_return_via_cx_bnd): Ditto.
1185 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
1186 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
1187 (output_indirect_thunk): Ditto. Remove need_prefix argument.
1188 (output_indirect_thunk_function): Remove handling of
1189 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
1190 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
1191 (ix86_save_reg): Remove handling of crtl->return_bnd.
1192 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
1193 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
1194 and UNSPEC_BNDLX_ADDR.
1195 (ix86_output_indirect_branch_via_reg): Remove handling of
1196 indirect_thunk_prefix_bnd.
1197 (ix86_output_indirect_branch_via_push): Ditto.
1198 (ix86_output_function_return): Ditto.
1199 (ix86_output_indirect_function_return): Ditto.
1200 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
1201 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
1202 (CALL_USED_REGISTERS): Ditto.
1203 (REG_ALLOC_ORDER): Update for removal of bound registers.
1204 (HI_REGISTER_NAMES): Ditto.
1205 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
1206 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
1207 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
1208 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
1209 (FIRST_PSEUDO_REG): Update.
1210 (BND): Remove mode iterator.
1211 * config/i386/predicates.md (bnd_mem_operator): Remove.
1213 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
1215 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
1218 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
1220 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
1221 the case in which the permute needs only a single element and
1222 repeats for every vector of the result. Extend that case to
1223 handle variable-length vectors.
1224 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
1226 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
1229 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
1230 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
1232 2018-08-24 Richard Biener <rguenther@suse.de>
1234 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
1235 bb_flags_allocated members.
1236 (auto_flag): New RAII class for allocating flags.
1237 (auto_edge_flag): New RAII class for allocating edge flags.
1238 (auto_bb_flag): New RAII class for allocating bb flags.
1239 * cfgloop.c (verify_loop_structure): Allocate temporary edge
1241 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
1242 in favor of temporarily allocated BB flag.
1243 * hsa-brig.c: Re-order includes.
1244 * hsa-dump.c: Likewise.
1245 * hsa-regalloc.c: Likewise.
1246 * print-rtl.c: Likewise.
1247 * profile-count.c: Likewise.
1249 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
1252 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
1255 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
1258 * wide-int-range.cc (wide_int_range_div): Do not ignore result
1259 from wide_int_range_multiplicative_op.
1261 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1263 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
1266 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
1268 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
1271 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
1273 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
1274 full GENERIC expression used for replacement.
1276 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
1278 * tree-vrp.c (abs_extent_range): Remove.
1279 (extract_range_into_wide_ints): Pass wide ints by reference.
1280 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
1281 Pass wide ints by reference in all calls to
1282 extract_range_into_wide_ints.
1283 * wide-int-range.cc (wide_int_range_div): New.
1284 * wide-int-range.h (wide_int_range_div): New.
1285 (wide_int_range_includes_zero_p): New.
1286 (wide_int_range_zero_p): New.
1288 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
1290 * config/aarch64/aarch64.md (arches): New enum.
1291 (arch): New enum attr.
1292 (arch_enabled): New attr.
1293 (enabled): Now uses arch_enabled only.
1294 (simd, sve, fp16): Removed attribute.
1295 (fp): Attr now defined in terms of 'arch'.
1296 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
1297 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
1298 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
1299 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
1300 attributes into 'arch'.
1301 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
1302 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
1303 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
1304 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
1305 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
1306 'simd' attribute into 'arch'.
1307 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
1308 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
1309 Convert use of 'fp' attribute to 'arch'.
1310 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
1311 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
1313 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
1314 (different modes) Merge 'fp' and 'simd' into 'arch'.
1315 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
1318 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
1320 PR rtl-optimization/87026
1321 * expmed.c (canonicalize_comparison): If we can no longer create
1322 pseudoregisters, don't.
1324 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
1327 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
1329 * config/arm/arm.c (speculation_barrier_libfunc): New static
1331 (arm_init_libfuncs): Initialize it.
1332 (arm_emit_speculation_barrier): New function.
1333 * config/arm/arm.md (speculation_barrier): Call
1334 arm_emit_speculation_barrier for architectures that do not have
1336 (speculation_barrier_insn): Only match on Armv7 or later.
1338 2018-08-23 Richard Biener <rguenther@suse.de>
1341 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
1344 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
1346 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
1347 of single-vector TBLs.
1348 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
1351 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
1354 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
1355 aarch64_evpc_tbl guard.
1357 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
1359 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
1362 2018-08-22 Martin Sebor <msebor@redhat.com>
1365 * tree-pretty-print.c (pretty_print_string): Add argument.
1366 (dump_generic_node): Call to pretty_print_string with string size.
1368 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
1370 PR rtl-optimization/86771
1371 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
1372 of two SETs into those two SETs, one to be placed at i2, if that SETs
1373 destination is modified between i2 and i3.
1375 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1377 PR tree-optimization/86725
1378 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
1380 (vect_analyze_scalar_cycles_1): Check it.
1382 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1384 PR tree-optimization/86725
1385 * tree-vect-loop.c (vect_is_simple_reduction): When treating
1386 an outer loop phi as a double reduction, make sure that the
1387 single user of the phi result is an inner loop phi.
1389 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1391 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
1392 grouped stores with gaps to a strided group.
1394 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1396 * tree-vect-stmts.c (get_group_load_store_type)
1397 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
1398 first statement in a group.
1400 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1402 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
1403 the sequence used in gcc/gcc.c.
1405 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1408 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
1411 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1413 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
1414 Darwin10-specific unwinder-shim.
1415 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
1416 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
1417 New to cater for Darwin10 Rosetta.
1419 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1421 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
1422 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
1424 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1429 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
1430 (dwarf2out_switch_text_section): Generate a local label for the
1431 second function sub-section and apply it as the second FDE start
1433 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
1434 second sub-section start.
1436 2018-08-22 Richard Biener <rguenther@suse.de>
1438 PR tree-optimization/86988
1439 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
1441 2018-08-22 Richard Biener <rguenther@suse.de>
1443 PR tree-optimization/86945
1444 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
1446 2018-08-22 Alexandre Oliva <oliva@adacore.com>
1448 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
1449 a comment about how uses of r2 for .sdata2 come about.
1451 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
1453 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
1455 2018-08-21 Marek Polacek <polacek@redhat.com>
1457 PR c++/86981, Implement -Wpessimizing-move.
1458 * doc/invoke.texi: Document -Wpessimizing-move.
1460 2018-08-21 Jan Hubicka <jh@suse.cz>
1462 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
1463 * tree.h (is_redundant_typedef): Remove.
1464 * dwarf2out.c (is_redundant_typedef): Turn into static function.
1466 2018-08-21 Jan Hubicka <jh@suse.cz>
1468 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
1471 2018-08-21 Tamar Christina <tamar.christina@arm.com>
1473 * expmed.c (extract_low_bits): Reject invalid subregs early.
1475 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
1478 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
1481 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1483 * config/vxworks.h: Guard vxworks_asm_out_constructor and
1484 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
1485 * config/vxworks.c: Likewise.
1487 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1489 * config/vxworks.c: Set targetm.have_ctors_dtors
1490 if HAVE_INITFINI_ARRAY_SUPPORT.
1491 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
1492 if HAVE_INITFINI_ARRAY_SUPPORT.
1494 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1496 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
1497 default search path for VxWorks < 7.
1499 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
1501 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
1502 (get_format_string): Refer to c_getstr.
1504 2018-08-21 Tom de Vries <tdevries@suse.de>
1506 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
1507 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
1508 (debuginfo_early_stop): Declare.
1509 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
1510 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
1511 (debuginfo_early_stop): New function.
1512 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
1513 and debuginfo_early_stop.
1514 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
1515 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
1516 (general_init): Call debuginfo_early_init.
1517 (finalize): Call debuginfo_fini.
1518 (do_compile): Call debuginfo_init.
1519 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
1521 (@item -fdump-debug, @item -fdump-earlydebug): Add.
1523 2018-08-21 Tom de Vries <tdevries@suse.de>
1525 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
1526 flag_dump_noaddr and flag_dump_unnumbered.
1528 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
1530 * wide-int-range.cc (wide_int_range_abs): New.
1531 (wide_int_range_order_set): Rename from wide_int_range_min_max.
1532 * wide-int-range.h (wide_int_range_abs): New.
1533 (wide_int_range_min_max): New.
1534 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
1535 case to call wide_int_range_abs.
1536 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
1537 (extract_range_from_abs_expr): Delete.
1539 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
1542 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
1543 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
1546 2018-08-20 Nathan Sidwell <nathan@acm.org>
1547 Jeff Law <law@redhat.com>
1549 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
1550 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
1552 2018-08-20 David Malcolm <dmalcolm@redhat.com>
1555 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
1556 (decl_attributes): Likewise.
1557 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
1559 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1560 * diagnostic-core.h (class auto_diagnostic_group): New class.
1561 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
1562 (diagnostic_report_diagnostic): Handle the first diagnostics within
1564 (emit_diagnostic): Add auto_diagnostic_group instance.
1566 (inform_n): Likewise.
1567 (warning): Likewise.
1568 (warning_at): Likewise.
1569 (warning_n): Likewise.
1570 (pedwarn): Likewise.
1571 (permerror): Likewise.
1573 (error_n): Likewise.
1574 (error_at): Likewise.
1576 (fatal_error): Likewise.
1577 (internal_error): Likewise.
1578 (internal_error_no_backtrace): Likewise.
1579 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
1580 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
1581 * diagnostic.h (struct diagnostic_context): Add fields
1582 "diagnostic_group_nesting_depth",
1583 "diagnostic_group_emission_count", "begin_group_cb",
1585 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
1586 Add auto_diagnostic_group instance(s).
1587 (find_explicit_erroneous_behavior): Likewise.
1588 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
1589 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
1590 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
1591 (gimplify_va_arg_expr): Likewise.
1592 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
1593 (HSA_SORRY_AT): Likewise.
1594 * ipa-devirt.c (compare_virtual_tables): Likewise.
1595 (warn_odr): Likewise.
1596 * multiple_target.c (expand_target_clones): Likewise.
1597 * opts-common.c (cmdline_handle_error): Likewise.
1598 * reginfo.c (globalize_reg): Likewise.
1599 * substring-locations.c (format_warning_n_va): Likewise.
1600 * tree-inline.c (expand_call_inline): Likewise.
1601 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
1602 * tree-ssa-loop-niter.c
1603 (do_warn_aggressive_loop_optimizations): Likewise.
1604 * tree-ssa-uninit.c (warn_uninit): Likewise.
1605 * tree.c (warn_deprecated_use): Likewise.
1607 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
1610 * config/i386/i386.md (eh_return): Always update EH return
1611 address in word_mode.
1613 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
1615 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
1616 TARGET_SPLIT_COMPLEX_ARG is defined.
1618 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
1620 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
1622 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
1625 * expr.c (expand_assignment): Assert that bitpos is positive.
1626 (store_field): Likewise
1627 (expand_expr_real_1): Make sure that bitpos is positive.
1628 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
1631 2018-08-20 Nathan Sidwell <nathan@acm.org>
1633 * Makefile.in (CPP_ID_DATA_H): Delete.
1634 (CPP_INTERNAL_H): Don't add it.
1635 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
1636 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
1638 2018-08-20 Richard Biener <rguenther@suse.de>
1640 PR tree-optimization/78655
1641 * tree-vrp.c (extract_range_from_binary_expr_1): Make
1642 pointer + offset nonnull if either operand is nonnull work.
1644 2018-08-20 Tom de Vries <tdevries@suse.de>
1646 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
1647 unless the referenced die describes the added property using
1648 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
1649 Otherwise, add a DW_AT_location to the referenced die.
1651 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
1654 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
1655 register_operand when calling ix86_set_reg_reg_cost.
1656 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
1657 Set *total to 0 for operands that satisfy x86_64_immediate_operand
1658 predicate and to 1 otherwise.
1660 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
1662 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
1663 emit a diagnostic that it is not supported and reset the option.
1664 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
1665 supported and consume the option. (ASM_FINAL_SPEC): New.
1667 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
1669 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
1672 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
1674 C-SKY port: Documentation
1676 * doc/extend.texi (C-SKY Function Attributes): New section.
1677 * doc/invoke.texi (Option Summary): Add C-SKY options.
1678 (C-SKY Options): New section.
1679 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
1681 2018-08-17 Jojo <jijie_rong@c-sky.com>
1682 Huibin Wang <huibin_wang@c-sky.com>
1683 Sandra Loosemore <sandra@codesourcery.com>
1684 Chung-Lin Tang <cltang@codesourcery.com>
1686 C-SKY port: Backend implementation
1688 * config/csky/*: New.
1689 * common/config/csky/*: New.
1691 2018-08-17 Jojo <jijie_rong@c-sky.com>
1692 Huibin Wang <huibin_wang@c-sky.com>
1693 Sandra Loosemore <sandra@codesourcery.com>
1694 Chung-Lin Tang <cltang@codesourcery.com>
1695 Andrew Jenner <andrew@codesourcery.com>
1697 C-SKY port: Configury
1699 * config.gcc (csky-*-*): New.
1700 * configure.ac: Add csky to targets for dwarf2 debug_line support.
1701 * configure: Regenerated.
1703 2018-08-17 David Malcolm <dmalcolm@redhat.com>
1705 * dump-context.h: Include "dumpfile.h".
1706 (dump_context::dump_printf_va): Convert final param from va_list
1707 to va_list *. Convert from ATTRIBUTE_PRINTF to
1708 ATTRIBUTE_GCC_DUMP_PRINTF.
1709 (dump_context::dump_printf_loc_va): Likewise.
1710 * dumpfile.c: Include "stringpool.h".
1711 (make_item_for_dump_printf_va): Delete.
1712 (make_item_for_dump_printf): Delete.
1713 (class dump_pretty_printer): New class.
1714 (dump_pretty_printer::dump_pretty_printer): New ctor.
1715 (dump_pretty_printer::emit_items): New member function.
1716 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
1718 (dump_pretty_printer::emit_item): New member function.
1719 (dump_pretty_printer::stash_item): New member function.
1720 (dump_pretty_printer::format_decoder_cb): New member function.
1721 (dump_pretty_printer::decode_format): New member function.
1722 (dump_context::dump_printf_va): Reimplement in terms of
1723 dump_pretty_printer.
1724 (dump_context::dump_printf_loc_va): Convert final param from va_list
1726 (dump_context::begin_scope): Reimplement call to
1727 make_item_for_dump_printf.
1728 (dump_printf): Update for change to dump_printf_va.
1729 (dump_printf_loc): Likewise.
1730 (selftest::test_capture_of_dump_calls): Convert "stmt" from
1731 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
1732 with %T, %E, and %G.
1733 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
1734 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
1735 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
1736 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
1737 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
1738 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
1739 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
1740 within a dump_printf_loc call to "%wu".
1741 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
1742 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
1743 missing space after "=".
1744 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
1745 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
1746 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
1747 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
1748 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
1749 duplicate "vectorized" from message.
1751 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
1753 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
1754 polyNxK_t element's TYPE_STRING_FLAG.
1756 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
1758 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
1759 (they were unnamed before). Fix comments.
1761 2018-08-17 Nathan Sidwell <nathan@acm.org>
1763 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
1765 2018-08-17 Richard Biener <rguenther@suse.de>
1767 PR tree-optimization/86841
1768 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
1770 2018-08-17 Martin Liska <mliska@suse.cz>
1772 * common.opt: Remove Warn, Init and Report for options with
1773 Ignore/Deprecated flag. Warning is done automatically for
1775 * config/i386/i386.opt: Likewise.
1776 * config/ia64/ia64.opt: Likewise.
1777 * config/rs6000/rs6000.opt: Likewise.
1778 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
1779 Remove usage of flag_check_pointer_bounds.
1780 * lto-wrapper.c (merge_and_complain): Do not handle
1781 OPT_fcheck_pointer_bounds.
1782 (append_compiler_options): Likewise.
1783 * opt-functions.awk: Do not handle Deprecated.
1784 * optc-gen.awk: Check that Var, Report and Init are not
1785 used for an option with Ignore/Deprecated flag.
1786 * opts-common.c (decode_cmdline_option): Do not report
1788 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
1790 * opts.h (struct cl_option): Remove cl_deprecated flag.
1791 (CL_ERR_DEPRECATED): Remove error enum value.
1793 2018-08-17 Richard Biener <rguenther@suse.de>
1796 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
1797 across a va-arg-pack using call adjust its return value accordingly.
1799 2018-08-16 Martin Sebor <msebor@redhat.com>
1801 PR tree-optimization/86853
1802 * gimple-ssa-sprintf.c (struct format_result): Rename member.
1803 (struct fmtresult): Add member and initialize it in ctors.
1804 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
1805 (format_string): Handle %S the same as %ls. Set MAYFAIL.
1806 (format_directive): Set POSUNDER4K when MAYFAIL is set.
1807 (parse_directive): Handle %C same as %c.
1808 (sprintf_dom_walker::compute_format_length): Adjust.
1809 (is_call_safe): Adjust.
1811 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
1813 * builtins.c (c_strlen): Add new parameter eltsize. Use it
1814 for determining how to count the elements.
1815 * builtins.h (c_strlen): Adjust prototype.
1816 * expr.c (string_constant): Add new parameter mem_size.
1817 Set *mem_size appropriately.
1818 * expr.h (string_constant): Adjust protoype.
1819 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
1820 * gimple-fold.h (get_range_strlen): Adjust prototype.
1821 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
1822 (format_string): Call get_string_length with eltsize.
1824 2018-08-16 David Malcolm <dmalcolm@redhat.com>
1826 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
1827 to emit the span, rather than setting it as the prefix.
1829 2018-08-16 David Malcolm <dmalcolm@redhat.com>
1831 * diagnostic-show-locus.c (layout::start_annotation_line): Add
1832 "margin_char" parameter, defaulting to space. Use it in place
1833 of pp_space for the initial part of the margin.
1834 (layout::print_leading_fixits): Use '+' when filling the margin
1835 of line-insertion fix-it hints.
1837 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1839 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
1842 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1844 * config/rs6000/altivec.md: Don't set length attribute to the default
1846 * config/rs6000/darwin.md: Ditto.
1847 * config/rs6000/dfp.md: Ditto.
1848 * config/rs6000/htm.md: Ditto.
1849 * config/rs6000/rs6000.md: Ditto.
1850 * config/rs6000/sync.md: Ditto.
1851 * config/rs6000/vsx.md: Ditto.
1853 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1855 * config/rs6000/altivec.md: Don't set length attribute to the default
1856 value, for branch instructions.
1857 * config/rs6000/darwin.md: Ditto.
1858 * config/rs6000/rs6000.md: Ditto.
1860 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1862 * config/rs6000/rs6000.md (length): Always define as const_int 4.
1863 (unnamed conditional branch define_insn): Set length to 4 or 8
1864 depending on offset.
1865 (<bd>_<mode>): Similar, for alternative 0.
1866 (<bd>tf_<mode>): Ditto.
1868 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1870 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
1872 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
1874 * doc/rtl.texi: Replace old RTX class names with new names.
1877 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
1879 * expmed.h (canonicalize_comparison): New declaration.
1880 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
1881 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
1882 * optabs.c (prepare_cmp_insn): Likewise.
1883 * rtl.h (unsigned_condition_p): New function which checks if a
1884 comparison operator is unsigned.
1886 2018-08-16 Nathan Sidwell <nathan@acm.org>
1888 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
1889 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
1891 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1894 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
1895 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
1896 (mov<mov>): ..this and enable unconditionally.
1898 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1900 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
1902 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
1904 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
1905 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
1906 "Common" with "Target".
1908 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
1910 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
1911 * doc/invoke.texi (mmitigate-rop): Remove.
1912 * config/i386/i386.c: Do not include "regrename.h".
1913 (ix86_rop_should_change_byte_p, reg_encoded_number)
1914 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
1916 (ix86_reorg): Remove call to ix86_mitigate_rop.
1917 * config/i386/i386.md (attr "modrm_class"): Remove.
1918 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
1919 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
1920 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
1922 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
1924 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
1925 allow folding of mergeh() and mergel() for the float and double types.
1926 (fold_mergehl_helper): Rework to handle building a permute tree
1929 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
1931 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
1934 2018-08-15 David Malcolm <dmalcolm@redhat.com>
1936 * common.opt (fdiagnostics-show-labels): New option.
1937 * diagnostic-show-locus.c (class layout_range): Add field
1939 (class layout): Add field "m_show_labels_p".
1940 (layout_range::layout_range): Add param "label" and use it to
1942 (make_range): Pass in NULL for new "label" param of layout_range's
1944 (layout::layout): Initialize m_show_labels_p.
1945 (layout::maybe_add_location_range): Pass in loc_range->m_label
1946 when constructing layout_range instances.
1947 (struct line_label): New struct.
1948 (layout::print_any_labels): New member function.
1949 (layout::print_line): Call it if label-printing is enabled.
1950 (selftest::test_one_liner_labels): New test.
1951 (selftest::test_diagnostic_show_locus_one_liner): Call it.
1952 * diagnostic.c (diagnostic_initialize): Initialize
1953 context->show_labels_p.
1954 * diagnostic.h (struct diagnostic_context): Add field
1956 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
1957 -fno-diagnostics-show-labels.
1958 * dwarf2out.c (gen_producer_string): Add
1959 OPT_fdiagnostics_show_labels to the ignored options.
1960 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
1962 (gcc_rich_location::maybe_add_expr): Likewise.
1963 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
1964 label" param, defaulting to NULL.
1965 (gcc_rich_location::add_expr): Add "label" param.
1966 (gcc_rich_location::maybe_add_expr): Likewise.
1967 (class text_range_label): New class.
1968 (class range_label_for_type_mismatch): New class.
1969 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
1970 of format_warning_va.
1971 (fmtwarn_n): Likewise for new params of format_warning_n_va.
1972 * lto-wrapper.c (merge_and_complain): Add
1973 OPT_fdiagnostics_show_labels to the "pick one setting" options.
1974 (append_compiler_options): Likewise to the dropped options.
1975 (append_diag_options): Likewise to the passed-on options.
1976 * opts.c (common_handle_option): Handle the new option.
1977 * selftest-diagnostic.c
1978 (test_diagnostic_context::test_diagnostic_context): Enable
1980 * substring-locations.c: Include "gcc-rich-location.h".
1981 (format_warning_n_va): Add "fmt_label" and "param_label" params
1982 and use them as appropriate.
1983 (format_warning_va): Add "fmt_label" and "param_label" params,
1984 passing them on to format_warning_n_va.
1985 (format_warning_at_substring): Likewise.
1986 (format_warning_at_substring_n): Likewise.
1987 * substring-locations.h (format_warning_va): Add "fmt_label" and
1988 "param_label" params.
1989 (format_warning_n_va): Likewise.
1990 (format_warning_at_substring): Likewise.
1991 (format_warning_at_substring_n): Likewise.
1992 * toplev.c (general_init): Initialize global_dc->show_labels_p.
1994 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
1997 * builtins.c (expand_builtin_memcmp): Do not expand the call
1998 when overflow is detected.
2000 2018-08-15 Martin Sebor <msebor@redhat.com>
2002 PR tree-optimization/71625
2003 * config/aarch64/aarch64-builtins.c
2004 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
2006 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
2008 * config/s390/s390.c (s390_reorg): Remove loop.
2010 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
2013 (darwin_function_switched_text_sections): Delete.
2014 * gcc/config/darwin.h
2015 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
2017 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
2020 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
2021 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
2022 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
2024 2018-08-15 Martin Liska <mliska@suse.cz>
2026 PR tree-optimization/86925
2027 * predict.c (expr_expected_value_1): When taking
2028 later predictor, assign also probability.
2029 Use fold_build2_initializer_loc in order to fold
2030 the expression in -frounding-math.
2032 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
2034 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
2036 (expand_vec_perm_1): Try the new method.
2038 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
2041 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
2042 Check whether lra_live_max_point is 0 before dividing.
2044 2018-08-14 Martin Sebor <msebor@redhat.com>
2046 PR tree-optimization/86650
2047 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
2048 (vrp_prop::check_mem_ref): Same.
2050 2018-08-13 Liu Hao <lh_mouse@126.com>
2052 * pretty-print.c (eat_esc_sequence): Swap the foreground and
2053 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
2054 and clear it thereafter, as it only works for DBCS.
2056 2018-08-13 Liu Hao <lh_mouse@126.com>
2058 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
2059 handle returned by _get_osf_handle().
2061 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
2063 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
2064 for folding vec_perm.
2066 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
2068 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
2069 Add support for gimple-folding of vec_pack() and vec_unpack()
2072 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
2074 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
2075 vec_xst variants to the list.
2076 (rs6000_gimple_fold_builtin): Add support for folding unaligned
2077 vector loads and stores.
2079 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
2081 * config.gcc (rs6000-ibm-aix4.x): Delete.
2082 (rs6000-ibm-aix5.1): Delete.
2083 (rs6000-ibm-aix5.2): Delete.
2084 (rs6000-ibm-aix5.3): Delete.
2085 * config/rs6000/aix43.h: Delete.
2086 * config/rs6000/aix51.h: Delete.
2087 * config/rs6000/aix52.h: Delete.
2088 * config/rs6000/t-aix43: Delete.
2090 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
2092 * config/s390/s390.c (s390_decompose_constant_pool_ref):
2094 (s390_decompose_address): Factor out constant pool ref
2097 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2099 * config/nds32/nds32-predicates.c
2100 (nds32_can_use_bclr_p): Change return type as bool.
2101 (nds32_can_use_bset_p): Ditto.
2102 (nds32_can_use_btgl_p): Ditto.
2103 (nds32_can_use_bitci_p): Ditto.
2104 * config/nds32/nds32-protos.h
2105 (nds32_can_use_bclr_p): Change declaration.
2106 (nds32_can_use_bset_p): Ditto.
2107 (nds32_can_use_btgl_p): Ditto.
2108 (nds32_can_use_bitci_p): Ditto.
2110 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2112 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
2113 Support -msched-prolog-epilog option.
2114 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
2116 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2118 * common/config/nds32/nds32-common.c
2119 (nds32_option_optimization_table): Enalbe -malways-align.
2121 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2123 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
2125 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
2126 OPT_misr_secure_ case.
2127 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
2128 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
2129 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
2131 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
2132 (nds32_isr_info): New field security_level.
2133 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
2134 * config/nds32/nds32.md (return_internal): Consider critical attribute.
2135 * config/nds32/nds32.opt (misr-secure): New option.
2136 * config/nds32/nds32_init.inc: New file.
2137 * config/nds32/nds32_isr.h: New file.
2139 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
2141 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
2142 Update comment for atomic instructions.
2143 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
2144 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
2146 (atomic_loaddi): Revise fence expansion to only emit fence prior to
2147 load for __ATOMIC_SEQ_CST model.
2148 (atomic_loaddi_1): Remove float register target.
2149 (atomic_storedi): Handle CONST_INT values.
2150 (atomic_storedi_1): Remove float register source. Add special case
2152 (memory_barrier): New expander and insn.
2154 2018-08-11 Jakub Jelinek <jakub@redhat.com>
2156 PR tree-optimization/86835
2157 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
2158 new_stmt after def_gsi, make sure to insert new_square_stmt after
2159 that stmt, not 2 stmts before it.
2161 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
2164 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
2167 2018-08-10 Martin Liska <mliska@suse.cz>
2170 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
2172 * builtins.c (expand_builtin_expect_with_probability):
2174 (expand_builtin_expect_with_probability): New function.
2175 (build_builtin_expect_predicate): Add new argumnet probability
2176 for BUILT_IN_EXPECT_WITH_PROBABILITY.
2177 (fold_builtin_expect):
2180 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
2181 * builtins.h (fold_builtin_expect): Set new argument.
2182 * doc/extend.texi: Document __builtin_expect_with_probability.
2183 * doc/invoke.texi: Likewise.
2184 * gimple-fold.c (gimple_fold_call): Pass new argument.
2185 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
2186 also BUILT_IN_EXPECT_WITH_PROBABILITY.
2187 * predict.c (get_predictor_value): New function.
2188 (expr_expected_value): Add new argument probability. Assume
2189 that predictor and probability are always non-null.
2190 (expr_expected_value_1): Likewise. For __builtin_expect and
2191 __builtin_expect_with_probability set probability. Handle
2192 combination in binary expressions.
2193 (tree_predict_by_opcode): Simplify code by simply calling
2194 get_predictor_value.
2195 (pass_strip_predict_hints::execute): Add handling of
2196 BUILT_IN_EXPECT_WITH_PROBABILITY.
2197 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
2199 * tree.h (DECL_BUILT_IN_P): New function.
2201 2018-08-10 Martin Liska <mliska@suse.cz>
2203 PR tree-optimization/85799
2204 * passes.def: Add argument for pass_strip_predict_hints.
2205 * predict.c (class pass_strip_predict_hints): Add new argument
2207 (strip_predictor_early): New function.
2208 (pass_strip_predict_hints::execute): Call the function to
2210 (strip_predict_hints): New function.
2211 * predict.def: Fix comment.
2213 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
2215 * Makefile.in: Clarify which tm.texi to copy over to assert the
2216 right to grant a GFDL license for all.
2218 2018-08-09 Jeff Law <law@redhat.com>
2220 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
2223 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
2225 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
2228 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2230 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
2231 reductions for variable-length vectors.
2233 2018-08-09 David Malcolm <dmalcolm@redhat.com>
2236 * common.opt (fdiagnostics-show-line-numbers): New option.
2237 * diagnostic-show-locus.c (class layout): Add fields
2238 "m_show_line_numbers_p" and "m_linenum_width";
2239 (num_digits): New function.
2240 (test_num_digits): New function.
2241 (layout::layout): Initialize new fields. Update m_x_offset
2242 logic to handle any left margin.
2243 (layout::print_source_line): Print line number when requested.
2244 (layout::start_annotation_line): New member function.
2245 (layout::print_annotation_line): Call it.
2246 (layout::print_leading_fixits): Likewise.
2247 (layout::print_trailing_fixits): Likewise. Update calls to
2248 move_to_column for new parameter.
2249 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
2250 it to potentially call start_annotation_line.
2251 (layout::show_ruler): Call start_annotation_line.
2252 (selftest::test_line_numbers_multiline_range): New selftest.
2253 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
2254 and selftest::test_line_numbers_multiline_range.
2255 * diagnostic.c (diagnostic_initialize): Initialize
2256 show_line_numbers_p.
2257 * diagnostic.h (struct diagnostic_context): Add field
2258 "show_line_numbers_p".
2259 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
2260 -fno-diagnostics-show-line-numbers.
2261 * dwarf2out.c (gen_producer_string): Add
2262 OPT_fdiagnostics_show_line_numbers to the ignored options.
2263 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
2264 one setting" options.
2265 (append_compiler_options): Likewise to the dropped options.
2266 (append_diag_options): Likewise to the passed-on options.
2267 * opts.c (common_handle_option): Handle the new option.
2268 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
2270 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
2272 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
2273 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
2274 third argument of type "const signed char" to descriptions of
2275 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
2276 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
2277 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
2278 __builtin_bcdsub_ov functions.
2280 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2282 PR tree-optimization/86858
2283 * tree-vect-loop.c (vect_is_simple_reduction): Restore
2284 flow_bb_inside_loop_p calls.
2286 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2288 PR tree-optimization/86871
2289 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
2290 instead of gimple_assign_lhs.
2292 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
2295 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
2296 register constraint to operand 0.
2297 (add<mode>3_carryinC): Likewise.
2298 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
2300 2018-08-09 Martin Liska <mliska@suse.cz>
2303 * common.opt: Remove extra line.
2305 2018-08-09 Martin Liska <mliska@suse.cz>
2307 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
2308 at the end of a line, make first letter capital and end up
2309 a sentence with a dot.
2310 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
2311 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
2312 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
2313 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
2314 (PARAM_MAX_ISL_OPERATIONS): Likewise.
2315 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
2316 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
2317 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
2318 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
2319 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
2320 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
2321 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
2322 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
2323 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
2324 (PARAM_TREE_REASSOC_WIDTH): Likewise.
2325 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
2326 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
2327 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
2329 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
2332 * config/s390/s390.c (s390_option_override_internal): Reduce the
2333 stack-clash-protection-probe-interval param if it would be too big
2336 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
2339 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
2340 * config/m68k/m68k.c (handle_move_double): Don't call
2341 m68k_final_prescan_insn.
2342 (m68k_adjust_decorated_operand): Renamed from
2343 m68k_final_prescan_insn, remove first and third operand and
2345 (print_operand): Call it.
2346 (print_operand_address): Call it.
2348 2018-08-08 Nathan Sidwell <nathan@acm.org>
2350 * diagnostic.c (diagnostic_report_current_module): Use
2351 linemap_included_from & linemap_included_from_linemap.
2353 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
2355 * config/aarch64/aarch64-cores.def: Add phecda core.
2356 * config/aarch64/aarch64-tune.md: Regenerate.
2357 * doc/invoke.texi: Add phecda core.
2359 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
2362 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
2364 * config/s390/s390.md ("movti"): Add more alternatives for
2365 constant to GPR copies.
2367 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
2369 * config/s390/s390.c: Fix whitespace damage throughout the file.
2370 * config/s390/s390.h: Likewise.
2371 * config/s390/tpf.h: Likewise.
2373 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
2375 * config/s390/s390.c (s390_loadrelative_operand_p):
2376 Remove TARGET_CPU_ZARCH usages.
2377 (s390_rtx_costs): Likewise.
2378 (s390_legitimate_constant_p): Likewise.
2379 (s390_cannot_force_const_mem): Likewise.
2380 (legitimate_reload_constant_p): Likewise.
2381 (s390_preferred_reload_class): Likewise.
2382 (legitimize_pic_address): Likewise.
2383 (legitimize_tls_address): Likewise.
2384 (s390_split_branches): Removed.
2385 (s390_add_execute): Removed.
2386 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
2387 (s390_mainpool_start): Likewise.
2388 (s390_mainpool_finish): Likewise.
2389 (s390_mainpool_cancel): Removed.
2390 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
2391 (s390_chunkify_cancel): Likewise.
2392 (s390_return_addr_rtx): Likewise.
2393 (s390_register_info): Remove split_branches_pending_p uages.
2394 (s390_optimize_register_info): Likewise.
2395 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
2396 split_branches_pending_p usages.
2397 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
2398 (s390_load_got): Likewise.
2399 (s390_expand_split_stack_prologue): Likewise.
2400 (output_asm_nops): Likewise.
2401 (s390_function_profiler): Likewise.
2402 (s390_emit_call): Likewise.
2403 (s390_conditional_register_usage): Likewise.
2404 (s390_optimize_prologue): Likewise.
2405 (s390_reorg): Remove TARGET_CPU_ZARCH and
2406 split_branches_pending_p usages.
2407 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
2409 (s390_output_indirect_thunk_function): Likewise.
2410 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
2411 (TARGET_CPU_ZARCH_P): Removed.
2412 (struct machine_function): Remove split_branches_pending_p.
2413 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
2415 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
2417 * common/config/s390/s390-common.c (processor_flags_table):
2419 * config.gcc: Remove with_arch/with_tune support.
2420 * config/s390/2064.md: Remove cpu attribute comparisons.
2421 * config/s390/driver-native.c (s390_host_detect_local_cpu):
2423 * config/s390/linux.h (ASM_SPEC):
2424 Remove -march support.
2425 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
2426 Use a table to get an arch level.
2427 * config/s390/s390-opts.h (enum processor_type):
2429 * config/s390/s390.c
2430 (processor_table): Remove entries, add arch_level values.
2431 (s390_issue_rate): Remove cases.
2432 (s390_option_override): Adjust
2433 s390_option_override_internal() call.
2434 (s390_option_override_internal): Remove deprecation warning.
2435 (s390_valid_target_attribute_tree): Adjust
2436 s390_option_override_internal() call.
2437 * config/s390/s390.h (struct s390_processor):
2438 Share with s390-c.c, add arch_level field.
2439 * config/s390/s390.md:
2440 Remove occurrences in cpu attribute.
2441 * config/s390/s390.opt: Remove -march/-mtune support.
2442 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
2443 * doc/invoke.texi: Remove deprecation warning.
2445 2018-08-08 Luis Machado <luis.machado@linaro.org>
2447 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
2449 (qdf24xx_tunings): Set vector cost structure to
2450 qdf24xx_vector_cost.
2452 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
2453 <register_sextend>: Set to 3.
2455 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
2458 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
2459 * config/aarch64/aarch64-simd.md
2460 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
2461 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
2462 (aarch64_frecpx<mode>): ...this new pattern.
2463 * config/aarch64/aarch64-simd-builtins.def: Remove comment
2464 about aarch64_frecp<FRECP:frecp_suffix><mode>.
2466 2018-08-07 Martin Liska <mliska@suse.cz>
2469 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
2470 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
2471 * predict.def (PRED_MALLOC_NONNULL): New predictor.
2472 * doc/extend.texi: Document that malloc attribute adds
2475 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
2478 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2479 Define to speculation_safe_value_not_needed.
2481 2018-08-06 Jeff Law <law@redhat.com>
2483 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
2484 the vr_values instance to cprop_into_stmt.
2485 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
2486 (cprop_operand): Also query EVRP to determine if OP is a constant.
2488 2018-08-06 Nathan Sidwell <nathan@acm.org>
2490 * diagnostic.c (diagnostic_report_current_module): Reroll
2491 included-at loop. Translate text.
2493 2018-08-06 David Malcolm <dmalcolm@redhat.com>
2495 * function-tests.c (selftest::test_expansion_to_rtl): Call
2496 free_after_compilation.
2498 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2500 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
2502 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
2504 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
2505 loops with memory block operations from getting unrolled.
2507 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
2510 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2511 Define to speculation_safe_value_not_needed.
2513 2018-08-06 Jeff Law <law@redhat.com>
2515 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
2518 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2521 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
2522 with all enabled __intN types.
2524 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
2526 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2528 * alias.c (record_set): Check for clobber high.
2529 * cfgexpand.c (expand_gimple_stmt): Likewise.
2530 * combine-stack-adj.c (single_set_for_csa): Likewise.
2531 * combine.c (find_single_use_1): Likewise.
2532 (set_nonzero_bits_and_sign_copies): Likewise.
2533 (get_combine_src_dest): Likewise.
2534 (is_parallel_of_n_reg_sets): Likewise.
2535 (try_combine): Likewise.
2536 (record_dead_and_set_regs_1): Likewise.
2537 (reg_dead_at_p_1): Likewise.
2538 (reg_dead_at_p): Likewise.
2539 * dce.c (deletable_insn_p): Likewise.
2540 (mark_nonreg_stores_1): Likewise.
2541 (mark_nonreg_stores_2): Likewise.
2542 * df-scan.c (df_find_hard_reg_defs): Likewise.
2543 (df_uses_record): Likewise.
2544 (df_get_call_refs): Likewise.
2545 * dwarf2out.c (mem_loc_descriptor): Likewise.
2546 * haifa-sched.c (haifa_classify_rtx): Likewise.
2547 * ira-build.c (create_insn_allocnos): Likewise.
2548 * ira-costs.c (scan_one_insn): Likewise.
2549 * ira.c (equiv_init_movable_p): Likewise.
2550 (rtx_moveable_p): Likewise.
2551 (interesting_dest_for_shprep): Likewise.
2552 * jump.c (mark_jump_label_1): Likewise.
2553 * postreload-gcse.c (record_opr_changes): Likewise.
2554 * postreload.c (reload_cse_simplify): Likewise.
2555 (struct reg_use): Add source expr.
2556 (reload_combine): Check for clobber high.
2557 (reload_combine_note_use): Likewise.
2558 (reload_cse_move2add): Likewise.
2559 (move2add_note_store): Likewise.
2560 * print-rtl.c (print_pattern): Likewise.
2561 * recog.c (decode_asm_operands): Likewise.
2562 (store_data_bypass_p): Likewise.
2563 (if_test_bypass_p): Likewise.
2564 * regcprop.c (kill_clobbered_value): Likewise.
2565 (kill_set_value): Likewise.
2566 * reginfo.c (reg_scan_mark_refs): Likewise.
2567 * reload1.c (maybe_fix_stack_asms): Likewise.
2568 (eliminate_regs_1): Likewise.
2569 (elimination_effects): Likewise.
2570 (mark_not_eliminable): Likewise.
2571 (scan_paradoxical_subregs): Likewise.
2572 (forget_old_reloads_1): Likewise.
2573 * reorg.c (find_end_label): Likewise.
2574 (try_merge_delay_insns): Likewise.
2575 (redundant_insn): Likewise.
2576 (own_thread_p): Likewise.
2577 (fill_simple_delay_slots): Likewise.
2578 (fill_slots_from_thread): Likewise.
2579 (dbr_schedule): Likewise.
2580 * resource.c (update_live_status): Likewise.
2581 (mark_referenced_resources): Likewise.
2582 (mark_set_resources): Likewise.
2583 * rtl.c (copy_rtx): Likewise.
2584 * rtlanal.c (reg_referenced_p): Likewise.
2585 (single_set_2): Likewise.
2586 (noop_move_p): Likewise.
2587 (note_stores): Likewise.
2588 * sched-deps.c (sched_analyze_reg): Likewise.
2589 (sched_analyze_insn): Likewise.
2591 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2593 * cse.c (invalidate_reg): New function extracted from...
2594 (invalidate): ...here.
2595 (canonicalize_insn): Check for clobber high.
2596 (invalidate_from_clobbers): invalidate clobber highs.
2597 (invalidate_from_sets_and_clobbers): Likewise.
2598 (count_reg_usage): Check for clobber high.
2599 (insn_live_p): Likewise.
2600 * cselib.c (cselib_expand_value_rtx_1):Likewise.
2601 (cselib_invalidate_regno): Check for clobber in setter.
2602 (cselib_invalidate_rtx): Pass through setter.
2603 (cselib_invalidate_rtx_note_stores):
2604 (cselib_process_insn): Check for clobber high.
2605 * cselib.h (cselib_invalidate_rtx): Add operand.
2607 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2609 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
2610 (mark_not_eliminable): Likewise.
2611 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
2612 * lra-lives.c (process_bb_lives): Check for clobber high.
2613 * lra.c (new_insn_reg): Remember clobber highs.
2614 (collect_non_operand_hard_regs): Check for clobber high.
2615 (lra_set_insn_recog_data): Likewise.
2616 (add_regs_to_insn_regno_info): Likewise.
2617 (lra_update_insn_regno_info): Likewise.
2619 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2621 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
2622 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
2624 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2626 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
2627 (copy_insn_1): Likewise.
2628 (gen_hard_reg_clobber_high): New gen function.
2629 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
2630 * genemit.c (gen_exp): Likewise.
2631 (gen_emit_seq): Pass through info.
2632 (gen_insn): Check for CLOBBER_HIGH.
2633 (gen_expand): Pass through info.
2634 (gen_split): Likewise.
2635 (output_add_clobbers): Likewise.
2636 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
2637 (remove_clobbers): Likewise.
2638 * rtl.h (gen_hard_reg_clobber_high): New declaration.
2640 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2642 * doc/rtl.texi (clobber_high): Add.
2643 (parallel): Add in clobber high
2644 * rtl.c (rtl_check_failed_code3): Add function.
2645 * rtl.def (CLOBBER_HIGH): Add expression.
2646 * rtl.h (RTL_CHECKC3): Add macro.
2647 (rtl_check_failed_code3): Add declaration.
2648 (XC3EXP): Add macro.
2650 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
2653 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
2654 cfun->machine->max_used_stack_alignment if needed.
2656 2018-08-04 Martin Sebor <msebor@redhat.com>
2658 PR tree-optimization/86571
2659 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
2662 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
2665 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2668 2018-08-03 Jeff Law <law@redhat.com>
2671 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2672 Define to speculation_safe_value_not_needed.
2674 2018-08-03 David Malcolm <dmalcolm@redhat.com>
2676 * doc/gcov.texi (-x): Remove duplicate "to".
2677 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
2678 (-Wif-not-aligned): Remove duplicate "is".
2679 (-flto): Remove duplicate "the".
2680 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
2681 duplicate "v5.00.b".
2682 (MSP430 Options): Remove duplicate "and" from the description
2683 of "-mgprel-sec=regexp".
2684 (x86 Options): Remove duplicate copies of "vmldLog102" and
2685 vmlsLog104 from description of "-mveclibabi=type".
2687 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
2689 * internal-fn.h (first_commutative_argument): Declare.
2690 * internal-fn.c (first_commutative_argument): New function.
2691 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
2692 restrictions for pattern statements. Use first_commutative_argument
2693 to look for commutative operands in calls to internal functions.
2695 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
2697 * Makefile.in (wide-int-range.o): New.
2698 * tree-vrp.c: Move all the wide_int_* functions to...
2699 * wide-int-range.cc: ...here.
2700 * tree-vrp.h: Move all the wide_int_* prototypes to...
2701 * wide-int-range.h: ...here.
2703 2018-08-03 Tom de Vries <tdevries@suse.de>
2705 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
2707 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
2708 * except.c (output_function_exception_table): Do early exit if
2709 targetm_common.except_unwind_info (&global_options) == UI_NONE.
2711 2018-08-03 Martin Liska <mliska@suse.cz>
2713 * predict.c (dump_prediction): Change to 2 digits
2716 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
2718 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
2720 (falkor_am_1_gtov_gtov): ... a new insn reservation.
2722 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
2724 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
2725 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
2726 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
2727 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
2728 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
2730 2018-08-02 David Malcolm <dmalcolm@redhat.com>
2732 * diagnostic-show-locus.c (diagnostic_show_locus): Use
2733 pp_take_prefix when saving the existing prefix.
2734 * diagnostic.c (diagnostic_append_note): Likewise.
2735 * langhooks.c (lhd_print_error_function): Likewise.
2736 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
2737 param's type. Free the existing prefix.
2738 (pp_take_prefix): New function.
2739 (pretty_printer::pretty_printer): Drop the prefix parameter.
2740 Rename the length parameter to match the comment.
2741 (pretty_printer::~pretty_printer): Free the prefix.
2742 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
2744 (struct pretty_printer): Drop the "const" from "prefix" field's
2745 type and clarify memory management.
2746 (pp_set_prefix): Drop the "const" from the 2nd param.
2747 (pp_take_prefix): New decl.
2749 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
2751 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
2752 for word_mode_ok here instead of passing as argument.
2753 (expand_block_compare): Change select_block_compare_mode() call.
2754 (expand_strncmp_gpr_sequence): New function.
2755 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
2757 2018-08-02 Jeff Law <law@redhat.com>
2760 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2761 Define to speculation_safe_value_not_needed.
2764 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2765 Define to speculation_safe_value_not_needed.
2767 2018-08-02 Tom de Vries <tdevries@suse.de>
2770 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
2771 function. Return UI_TARGET unconditionally.
2772 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
2773 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
2775 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
2777 * genemit.c (print_overload_test): Fix typo.
2779 2018-08-02 Richard Biener <rguenther@suse.de>
2781 PR tree-optimization/86816
2782 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
2783 which checks for value availability before querying it.
2784 (gvn_uses_equal): Use it.
2785 (same_succ_hash): Likewise.
2786 (gimple_equal_p): Likewise.
2788 2018-08-02 Nick Clifton <nickc@redhat.com>
2791 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2792 Define to speculation_safe_value_not_needed.
2795 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2796 Define to speculation_safe_value_not_needed.
2799 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2800 Define to speculation_safe_value_not_needed.
2803 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2804 Define to speculation_safe_value_not_needed.
2807 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2808 Define to speculation_safe_value_not_needed.
2811 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2812 Define to speculation_safe_value_not_needed.
2815 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2816 Define to speculation_safe_value_not_needed.
2819 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2820 Define to speculation_safe_value_not_needed.
2823 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
2824 speculation_safe_value_not_needed.
2827 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
2828 to speculation_safe_value_not_needed.
2830 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
2832 * doc/md.texi: Expand the documentation of instruction names
2833 to mention port-local uses. Document '@' in pattern names.
2834 * read-md.h (overloaded_instance, overloaded_name): New structs.
2836 (md_reader::handle_overloaded_name): New member function.
2837 (md_reader::get_overloads): Likewise.
2838 (md_reader::m_first_overload): New member variable.
2839 (md_reader::m_next_overload_ptr): Likewise.
2840 (md_reader::m_overloads_htab): Likewise.
2841 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
2842 m_next_overload_ptr and m_overloads_htab.
2843 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
2844 (get_mode_token, get_code_token, get_int_token): New functions.
2845 (map_attr_string): Add an optional argument that passes back
2846 the associated iterator.
2847 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
2848 (md_reader::handle_overloaded_name, add_overload_instance): New
2850 (apply_iterators): Handle '@' names. Report an error if '@'
2851 is used without iterators.
2852 (initialize_iterators): Initialize the new iterator_group fields.
2853 * genopinit.c (handle_overloaded_code_for)
2854 (handle_overloaded_gen): New functions.
2855 (main): Use them to print declarations of maybe_code_for_* and
2856 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
2857 * genemit.c (print_overload_arguments, print_overload_test)
2858 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
2859 (main): Use it to print definitions of maybe_code_for_* and
2860 maybe_gen_* functions.
2861 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
2862 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
2863 instead of explicit mode checks.
2864 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
2865 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
2866 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
2867 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
2868 (aarch64_expand_compare_and_swap): Likewise
2869 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
2870 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
2871 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
2872 (aarch64_constant_pool_reload_icode): Delete.
2873 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
2874 instead of aarch64_constant_pool_reload_icode. Use
2875 code_for_aarch64_reload_mov instead of explicit mode checks.
2876 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
2877 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
2878 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
2879 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
2880 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
2881 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
2882 (aarch64_atomic_load_op_code): Delete.
2883 (aarch64_emit_atomic_load_op): Likewise.
2884 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
2885 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
2886 instead of aarch64_emit_atomic_load_op.
2887 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
2888 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
2889 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
2890 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
2891 character before the pattern name.
2892 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
2893 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
2894 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
2895 (aarch64_frecps<mode>): Likewise.
2896 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
2897 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
2898 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
2899 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
2900 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
2902 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
2904 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
2905 Allow HFmode constants if TARGET_FP_F16INST.
2907 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
2910 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
2911 No longer check last store for clobber of address register.
2913 2018-08-02 Martin Liska <mliska@suse.cz>
2915 PR gcov-profile/86817
2916 * gcov.c (process_all_functions): New function.
2918 (process_file): Move functions processing to
2919 process_all_functions.
2921 2018-08-02 David Malcolm <dmalcolm@redhat.com>
2923 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
2924 "const" to the "gimple *" and "rtx_insn *" parameters.
2925 * dumpfile.h (dump_user_location_t::dump_user_location_t):
2927 (dump_location_t::dump_location_t): Likewise.
2929 2018-08-01 Martin Sebor <msebor@redhat.com>
2931 PR tree-optimization/86650
2932 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
2933 rather than a "gcall *". Directly pass the data of interest
2934 to percent_K_format, rather than building a temporary CALL_EXPR
2936 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
2937 (gimple_fold_builtin_strncat): Adjust.
2938 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
2939 gcall* argument with gimple*.
2940 * gimple-ssa-warn-restrict.c (check_call): Same.
2941 (wrestrict_dom_walker::before_dom_children): Same.
2942 (builtin_access::builtin_access): Same.
2943 (check_bounds_or_overlap): Same
2944 (maybe_diag_overlap): Same.
2945 (maybe_diag_offset_bounds): Same.
2946 * tree-diagnostic.c (default_tree_printer): Move usage of
2947 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
2949 * tree-pretty-print.c (percent_K_format): Add argument.
2950 * tree-pretty-print.h: Add argument.
2951 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
2952 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
2953 (maybe_diag_stxncpy_trunc): Same.
2954 (handle_builtin_stxncpy): Same.
2955 (handle_builtin_strcat): Same.
2957 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2959 * match.pd: Optimise pointer range checks.
2961 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2963 PR tree-optimization/86758
2964 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
2965 to remove pattern statements.
2967 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2969 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
2970 result of dfs_enumerate_from when constructing stmt_vec_infos,
2971 instead of additionally calling get_loop_body.
2973 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2975 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
2977 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
2978 When creating an iv, assert that the step is not known to be zero.
2979 (vect_setup_realignment): Update call accordingly.
2980 * tree-vect-stmts.c (vectorizable_store): Likewise.
2981 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
2983 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2985 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
2986 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
2987 (vectorizable_reduction): Likewise.
2988 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2989 (vect_detect_hybrid_slp_stmts): Likewise.
2990 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2992 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
2994 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
2995 (wide_int_set_zero_nonzero_bits): ...this.
2996 (zero_nonzero_bits_from_vr): Rename to...
2997 (vrp_set_zero_nonzero_bits): ...this.
2998 (extract_range_from_multiplicative_op_1): Abstract wide int
3000 (wide_int_range_multiplicative_op): ...here.
3001 (extract_range_from_binary_expr_1): Extract wide int binary
3002 operations into their own functions.
3003 (wide_int_range_lshift): New.
3004 (wide_int_range_can_optimize_bit_op): New.
3005 (wide_int_range_shift_undefined_p): New.
3006 (wide_int_range_bit_xor): New.
3007 (wide_int_range_bit_ior): New.
3008 (wide_int_range_bit_and): New.
3009 (wide_int_range_trunc_mod): New.
3010 (extract_range_into_wide_ints): New.
3011 (vrp_shift_undefined_p): New.
3012 (extract_range_from_multiplicative_op): New.
3013 (vrp_can_optimize_bit_op): New.
3014 * tree-vrp.h (value_range::dump): New.
3015 (wide_int_range_multiplicative_op): New.
3016 (wide_int_range_lshift):New.
3017 (wide_int_range_shift_undefined_p): New.
3018 (wide_int_range_bit_xor): New.
3019 (wide_int_range_bit_ior): New.
3020 (wide_int_range_bit_and): New.
3021 (wide_int_range_trunc_mod): New.
3022 (zero_nonzero_bits_from_bounds): Rename to...
3023 (wide_int_set_zero_nonzero_bits): ...this.
3024 (zero_nonzero_bits_from_vr): Rename to...
3025 (vrp_set_zero_nonzero_bits): ...this.
3026 (range_easy_mask_min_max): Rename to...
3027 (wide_int_range_can_optimize_bit_op): this.
3028 * vr-values.c (simplify_bit_ops_using_ranges): Rename
3029 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
3031 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3033 * tree-vectorizer.h (vect_orig_stmt): New function.
3034 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
3035 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
3036 (vect_create_epilog_for_reduction): Likewise.
3037 (vectorizable_live_operation): Likewise.
3038 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
3039 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
3040 * tree-vect-stmts.c (vectorizable_call): Likewise.
3041 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
3043 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3045 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
3047 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
3048 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
3049 (vect_transform_loop): Likewise.
3050 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
3052 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3054 * tree-vectorizer.h (vect_schedule_slp): Return void.
3055 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
3056 (vect_schedule_slp): Likewise.
3058 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3060 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
3062 (vect_transform_loop): Update calls accordingly. Schedule SLP
3063 instances before the main loop, if any exist.
3065 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3067 PR tree-optimization/86749
3068 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
3069 If the lhs is used in a COND_EXPR, check that it is being used
3070 as the "then" or "else" value.
3072 2018-08-01 Tom de Vries <tdevries@suse.de>
3075 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
3076 speculation_safe_value_not_needed.
3078 2018-08-01 Richard Biener <rguenther@suse.de>
3080 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
3083 2018-08-01 Martin Liska <mliska@suse.cz>
3085 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
3086 format how successful transformation is dumped.
3087 (gimple_mod_pow2_value_transform): Likewise.
3088 (gimple_mod_subtract_transform): Likewise.
3089 (gimple_stringops_transform): Likewise.
3091 2018-08-01 Martin Liska <mliska@suse.cz>
3094 * value-prof.c (interesting_stringop_to_profile_p):
3095 Simplify the code and add BUILT_IN_MEMMOVE.
3096 (gimple_stringops_transform): Likewise.
3098 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
3100 * config/aarch64/aarch64-simd.md
3101 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
3102 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
3103 use GPI iterator instead of SI mode.
3105 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
3107 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
3108 rs6000_speculation_barrier.
3109 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
3110 new barrier pattern name.
3112 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
3114 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
3115 (speculation_barrier): New insn.
3117 2018-08-01 Richard Biener <rguenther@suse.de>
3120 * graphite.h: Include isl/id.h and isl/space.h to allow build
3123 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
3126 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3127 mode for COFF targets.
3128 * defaults.h (TARGET_COFF): Define.
3129 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
3130 TARGET_COFF): Define.
3131 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
3132 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
3133 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
3135 2018-07-31 Alexandre Oliva <oliva@adacore.com>
3137 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
3138 * gimple-streamer-out.c (output_bb): Save it.
3139 * lto-streamer-in.c (input_struct_function_base): Restore
3140 instance discriminator if available. Create map on demand.
3141 * lto-streamer-out.c (output_struct_function_base): Save it if
3143 * final.c (decl_to_instance_map): Document LTO strategy.
3145 2018-07-31 Alexandre Oliva <oliva@adacore.com>
3146 Olivier Hainque <hainque@adacore.com>
3148 * debug.h (decl_to_instance_map_t): New type.
3149 (decl_to_instance_map): Declare.
3150 (maybe_create_decl_to_instance_map): New inline function.
3151 * final.c (bb_discriminator, last_bb_discriminator): New statics,
3152 to track basic block discriminators.
3153 (final_start_function_1): Initialize them.
3154 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
3156 (decl_to_instance_map): New variable.
3157 (map_decl_to_instance, maybe_set_discriminator): New functions.
3158 (notice_source_line): Set discriminator.
3160 2018-07-31 Ian Lance Taylor <iant@golang.org>
3162 * targhooks.c (default_have_speculation_safe_value): Add
3165 2018-07-31 David Malcolm <dmalcolm@redhat.com>
3167 * dump-context.h: Include "pretty-print.h".
3168 (dump_context::refresh_dumps_are_enabled): New decl.
3169 (dump_context::emit_item): New decl.
3170 (class dump_context): Add fields "m_test_pp" and
3172 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
3173 (temp_dump_context::get_dumped_text): New decl.
3174 (class temp_dump_context): Add field "m_pp".
3175 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
3176 (dump_context::refresh_dumps_are_enabled): ...and add a test for
3178 (set_dump_file): Update for above change.
3179 (set_alt_dump_file): Likewise.
3180 (dump_loc): New overload, taking a pretty_printer *.
3181 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
3182 to any test pretty-printer.
3183 (make_item_for_dump_gimple_stmt): New function, adapted from
3184 optinfo::add_gimple_stmt.
3185 (dump_context::dump_gimple_stmt): Call it, and use the result,
3186 eliminating the direct usage of dump_file and alt_dump_file in
3187 favor of indirectly using them via emit_item.
3188 (make_item_for_dump_gimple_expr): New function, adapted from
3189 optinfo::add_gimple_expr.
3190 (dump_context::dump_gimple_expr): Call it, and use the result,
3191 eliminating the direct usage of dump_file and alt_dump_file in
3192 favor of indirectly using them via emit_item.
3193 (make_item_for_dump_generic_expr): New function, adapted from
3195 (dump_context::dump_generic_expr): Call it, and use the result,
3196 eliminating the direct usage of dump_file and alt_dump_file in
3197 favor of indirectly using them via emit_item.
3198 (make_item_for_dump_printf_va): New function, adapted from
3199 optinfo::add_printf_va.
3200 (make_item_for_dump_printf): New function.
3201 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
3202 and use the result, eliminating the direct usage of dump_file and
3203 alt_dump_file in favor of indirectly using them via emit_item.
3204 (make_item_for_dump_dec): New function.
3205 (dump_context::dump_dec): Call it, and use the result,
3206 eliminating the direct usage of dump_file and alt_dump_file in
3207 favor of indirectly using them via emit_item.
3208 (make_item_for_dump_symtab_node): New function, adapted from
3209 optinfo::add_symtab_node.
3210 (dump_context::dump_symtab_node): Call it, and use the result,
3211 eliminating the direct usage of dump_file and alt_dump_file in
3212 favor of indirectly using them via emit_item.
3213 (dump_context::begin_scope): Reimplement, avoiding direct usage
3214 of dump_file and alt_dump_file in favor of indirectly using them
3216 (dump_context::emit_item): New member function.
3217 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
3218 Set up test pretty-printer on the underlying context. Call
3219 refresh_dumps_are_enabled.
3220 (temp_dump_context::~temp_dump_context): Call
3221 refresh_dumps_are_enabled.
3222 (temp_dump_context::get_dumped_text): New member function.
3223 (selftest::verify_dumped_text): New function.
3224 (ASSERT_DUMPED_TEXT_EQ): New macro.
3225 (selftest::test_capture_of_dump_calls): Run all tests twice, with
3226 and then without optinfo enabled. Add uses of
3227 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
3228 * dumpfile.h: Update comment for the dump_* API.
3229 * optinfo-emit-json.cc
3230 (selftest::test_building_json_from_dump_calls): Update for new
3231 param for temp_dump_context ctor.
3232 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
3233 and "m_owned" field.
3234 (optinfo_item::~optinfo_item): Likewise.
3235 (optinfo::add_item): New member function.
3236 (optinfo::emit): Update comment.
3237 (optinfo::add_string): Delete.
3238 (optinfo::add_printf): Delete.
3239 (optinfo::add_printf_va): Delete.
3240 (optinfo::add_gimple_stmt): Delete.
3241 (optinfo::add_gimple_expr): Delete.
3242 (optinfo::add_tree): Delete.
3243 (optinfo::add_symtab_node): Delete.
3244 (optinfo::add_dec): Delete.
3245 * optinfo.h (class dump_context): New forward decl.
3246 (optinfo::add_item): New decl.
3247 (optinfo::add_string): Delete.
3248 (optinfo::add_printf): Delete.
3249 (optinfo::add_printf_va): Delete.
3250 (optinfo::add_gimple_stmt): Delete.
3251 (optinfo::add_gimple_expr): Delete.
3252 (optinfo::add_tree): Delete.
3253 (optinfo::add_symtab_node): Delete.
3254 (optinfo::add_dec): Delete.
3255 (optinfo::add_poly_int): Delete.
3256 (optinfo_item::optinfo_item): Remove "owned" param.
3257 (class optinfo_item): Remove field "m_owned".
3259 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3262 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
3263 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
3264 requested variable alignment.
3265 (expand_one_ssa_partition): Likewise.
3266 (expand_one_var): Likewise.
3268 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3270 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3271 to speculation_safe_value_not_needed.
3273 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3275 * targhooks.h (speculation_safe_value_not_needed): New prototype.
3276 * targhooks.c (speculation_safe_value_not_needed): New function.
3277 * target.def (have_speculation_safe_value): Update documentation.
3278 * doc/tm.texi: Regenerated.
3280 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3282 * config/aarch64/iterators.md (ALLI_TI): New iterator.
3283 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
3285 (despeculate_copy<ALLI:mode>_insn): New insn.
3286 (despeculate_copyti_insn): New insn.
3287 (despeculate_simple<ALLI:mode>): New insn
3288 (despeculate_simpleti): New insn.
3289 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
3291 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
3292 aarch64_speculation_safe_value.
3293 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
3295 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3297 * config/aarch64/aarch64-speculation.cc: New file.
3298 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
3299 before pass_reorder_blocks.
3300 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
3302 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
3303 X14 and X15 when tracking speculation.
3304 * config/aarch64/aarch64.md (register name constants): Add
3305 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
3306 (unspec): Add UNSPEC_SPECULATION_TRACKER.
3307 (speculation_barrier): New insn attribute.
3308 (cmp<mode>): Allow SP in comparisons.
3309 (speculation_tracker): New insn.
3310 (speculation_barrier): Add speculation_barrier attribute.
3311 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
3312 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
3313 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
3315 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3317 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
3318 aarch64_track_speculation is true.
3319 (tb<optab><mode>1): Likewise.
3320 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
3321 generate CB[N]Z when tracking speculation.
3322 (aarch64_split_compare_and_swap): Likewise.
3323 (aarch64_split_atomic_op): Likewise.
3325 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3327 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
3329 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3331 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
3332 (speculation_barrier): New insn.
3334 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3336 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
3337 * config/arm/arm.md (speculation_barrier): New expand.
3338 (speculation_barrier_insn): New pattern.
3340 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3342 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
3343 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
3344 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
3345 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
3347 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
3348 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
3349 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
3350 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
3351 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
3352 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
3353 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
3354 * builtins.c (expand_speculation_safe_value): New function.
3355 (expand_builtin): Call it.
3356 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
3357 * doc/extend.texi: Document __builtin_speculation_safe_value.
3358 * doc/md.texi: Document "speculation_barrier" pattern.
3359 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
3360 TARGET_HAVE_SPECULATION_SAFE_VALUE.
3361 * doc/tm.texi: Regenerated.
3362 * target.def (have_speculation_safe_value, speculation_safe_value): New
3364 * targhooks.c (default_have_speculation_safe_value): New function.
3365 (default_speculation_safe_value): New function.
3366 * targhooks.h (default_have_speculation_safe_value): Add prototype.
3367 (default_speculation_safe_value): Add prototype.
3369 2018-07-31 David Malcolm <dmalcolm@redhat.com>
3371 * dump-context.h (dump_context::dump_loc): New decl.
3372 * dumpfile.c (dump_context::dump_loc): New member function.
3373 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
3374 and dump_gimple_stmt.
3375 (dump_context::dump_gimple_expr_loc): Likewise, using
3377 (dump_context::dump_generic_expr_loc): Likewise, using
3379 (dump_context::dump_printf_loc_va): Likewise, using
3381 (dump_context::begin_scope): Explicitly using the global function
3382 "dump_loc", rather than the member function.
3384 2018-07-31 Martin Sebor <msebor@redhat.com>
3386 PR tree-optimization/86741
3387 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
3389 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
3391 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
3392 depend on whether prefetch instructions will be emitted or not.
3393 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
3394 will be emitted or not.
3395 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
3396 (TARGET_SETMEM_PFD): New macros.
3398 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3400 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
3401 (NULL_STMT_VEC_INFO): Delete.
3402 (stmt_vec_info::operator*): Likewise.
3403 (stmt_vec_info::operator gimple *): Likewise.
3404 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
3405 of NULL_STMT_VEC_INFO.
3406 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3407 (vect_reassociating_reduction_p): Likewise.
3408 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
3409 (vectorizable_store): Likewise.
3410 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
3411 (vec_info::free_stmt_vec_infos): Likewise.
3413 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3415 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
3416 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
3418 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3420 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
3421 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
3422 (vec_info::free_stmt_vec_info): New private member functions.
3423 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
3424 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
3425 * tree-parloops.c (gather_scalar_reductions): Remove calls to
3426 set_stmt_vec_info_vec and free_stmt_vec_infos.
3427 * tree-vect-loop.c (_loop_vec_info): Remove call to
3428 set_stmt_vec_info_vec.
3429 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
3430 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
3431 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
3432 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
3433 (vec_info::free_stmt_vec_info): ...these new functions. Remove
3434 assignments in {vec_info::,}new_stmt_vec_info that are redundant
3435 with the clearing in the xcalloc.
3436 (stmt_vec_info_vec): Delete.
3437 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
3438 (vectorize_loops): Likewise.
3439 (vec_info::~vec_info): Remove argument from call to
3440 free_stmt_vec_infos.
3441 (vec_info::add_stmt): Remove vinfo argument from call to
3444 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3446 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
3447 rather than a gimple stmt.
3448 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
3449 information for pattern statements when passed the original
3450 statement; instead wait to be passed the pattern statement itself.
3451 Don't call set_vinfo_for_stmt here.
3452 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
3453 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
3454 stmt_vec_infos here.
3455 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
3456 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
3457 stmt_vec_infos entry.
3459 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3461 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
3462 * tree-vectorizer.c (vec_info::replace_stmt): New function.
3463 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
3464 * tree-vect-stmts.c (vectorizable_call): Likewise.
3465 (vectorizable_simd_clone_call): Likewise.
3467 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3469 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
3470 * tree-vectorizer.c (vec_info::remove_stmt): New function.
3471 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
3472 * tree-vect-loop.c (vect_transform_loop): Likewise.
3473 * tree-vect-slp.c (vect_schedule_slp): Likewise.
3474 * tree-vect-stmts.c (vect_remove_stores): Likewise.
3476 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3478 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
3479 (vect_dr_stmt): Delete.
3480 * tree-vectorizer.c (vec_info::lookup_dr): New function.
3481 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
3483 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
3484 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
3485 (vect_verify_datarefs_alignment, vect_peeling_supportable)
3486 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
3487 (vect_analyze_data_refs): Likewise.
3488 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
3490 (vect_find_same_alignment_drs): Likewise.
3491 (vect_slp_analyze_node_dependences): Update calls accordingly.
3492 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
3493 instead of DR_VECT_AUX.
3494 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
3495 of a vector data references. Use vec_info::lookup_dr instead of
3497 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
3498 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
3499 instead of DR_VECT_AUX.
3501 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3503 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
3505 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
3507 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3508 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3509 (vect_gen_prolog_loop_niters): Likewise.
3511 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3513 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
3514 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
3515 (vect_known_alignment_in_bytes, vect_dr_behavior)
3516 (vect_get_scalar_dr_size): Take references as dr_vec_infos
3517 instead of data_references. Update calls to other routines for
3518 which the same change has been made.
3519 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
3520 dr_vec_infos instead of stmt_vec_infos.
3521 (vect_analyze_data_ref_dependence): Update call accordingly.
3522 (vect_slp_analyze_data_ref_dependence)
3523 (vect_record_base_alignments): Use DR_VECT_AUX.
3524 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
3525 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
3526 (vector_alignment_reachable_p, vect_get_data_access_cost)
3527 (vect_peeling_supportable, vect_analyze_group_access_1)
3528 (vect_analyze_group_access, vect_analyze_data_ref_access)
3529 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
3530 (vect_compile_time_alias, vect_small_gap_p)
3531 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
3532 (vect_supportable_dr_alignment): Take references as dr_vec_infos
3533 instead of data_references. Update calls to other routines for
3534 which the same change has been made.
3535 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
3536 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
3537 (vect_slp_analyze_and_verify_node_alignment)
3538 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
3539 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
3540 (vect_setup_realignment): Use dr_vec_infos. Update calls after
3542 (_vect_peel_info::dr): Replace with...
3543 (_vect_peel_info::dr_info): ...this new field.
3544 (vect_peeling_hash_get_most_frequent)
3545 (vect_peeling_hash_choose_best_peeling): Update accordingly.
3546 (vect_peeling_hash_get_lowest_cost):
3547 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
3548 routines for which the same change has been made.
3549 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
3551 * tree-vect-loop-manip.c (get_misalign_in_elems)
3552 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
3554 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3555 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
3556 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
3557 (get_group_load_store_type, get_negative_load_store_type)
3558 (vect_get_data_ptr_increment, vectorizable_store)
3559 (vectorizable_load): Likewise.
3560 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
3561 Update calls to other routines for which the same change has been made.
3563 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3565 * tree-vectorizer.h (vec_info::move_dr): New member function.
3566 (dataref_aux): Rename to...
3567 (dr_vec_info): ...this and add "dr" and "stmt" fields.
3568 (_stmt_vec_info::dr_aux): Update accordingly.
3569 (_stmt_vec_info::data_ref_info): Delete.
3570 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
3571 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
3572 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
3573 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
3575 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
3576 (STMT_VINFO_DR_INFO): New macro.
3577 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
3578 (set_dr_misalignment): Update after rename of dataref_aux.
3579 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
3580 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
3581 initialization of STMT_VINFO_DATA_REF.
3582 * tree-vectorizer.c (vec_info::move_dr): New function.
3583 * tree-vect-patterns.c (vect_recog_bool_pattern)
3584 (vect_recog_mask_conversion_pattern)
3585 (vect_recog_gather_scatter_pattern): Use it.
3586 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
3587 the "dr" and "stmt" fields of dr_vec_info instead of
3588 STMT_VINFO_DATA_REF.
3590 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3592 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
3593 (is_pattern_stmt_p): Use it.
3594 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
3595 on pattern statements.
3597 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3599 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
3600 original stmt as a stmt_vec_info rather than a gimple stmt.
3601 (vect_pattern_recog_1): Take the statement directly as a
3602 stmt_vec_info, rather than via a gimple_stmt_iterator.
3603 Update call to vect_mark_pattern_stmts.
3604 (vect_pattern_recog): Update calls accordingly.
3606 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3608 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
3609 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
3610 a vect_def_type for the first argument.
3611 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
3612 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
3613 operand if it isn't defined by a vectorized statement.
3614 (vect_build_gather_load_calls): Remove the mask_dt argument and
3615 update calls to vect_get_vec_def_for_stmt_copy.
3616 (vectorizable_bswap): Likewise the dt argument.
3617 (vectorizable_call): Update calls to vectorizable_bswap and
3618 vect_get_vec_def_for_stmt_copy.
3619 (vectorizable_simd_clone_call, vectorizable_assignment)
3620 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
3621 (vectorizable_comparison): Update calls to
3622 vect_get_vec_def_for_stmt_copy.
3623 (vectorizable_store): Likewise. Remove now-unnecessary calls to
3625 (vect_get_loop_based_defs): Remove dt argument and update call
3626 to vect_get_vec_def_for_stmt_copy.
3627 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
3628 and vect_get_vec_def_for_stmt_copy.
3629 (vectorizable_load): Update calls to vect_build_gather_load_calls
3630 and vect_get_vec_def_for_stmt_copy.
3631 * tree-vect-loop.c (vect_create_epilog_for_reduction)
3632 (vectorizable_reduction, vectorizable_live_operation): Update calls
3633 to vect_get_vec_def_for_stmt_copy.
3635 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3637 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
3638 and gimple stmt arguments with a stmt_vec_info.
3639 (vect_record_base_alignments): Update calls accordingly.
3640 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
3641 and gimple stmt arguments with a stmt_vec_info.
3642 (vect_build_slp_tree_1): Remove vinfo argument and update call
3643 to vect_record_max_nunits.
3644 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
3645 and vect_record_max_nunits.
3647 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3649 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
3650 file and take a stmt_vec_info instead of a gimple stmt.
3651 (supportable_widening_operation, vect_finish_replace_stmt)
3652 (vect_finish_stmt_generation, vect_get_store_rhs)
3653 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
3654 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
3655 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
3656 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
3657 (vect_create_data_ref_ptr, bump_vector_ptr)
3658 (vect_permute_store_chain, vect_setup_realignment)
3659 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
3660 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
3661 (vectorizable_reduction, vectorizable_induction)
3662 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
3663 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
3664 than gimple stmts as arguments.
3665 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
3666 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
3667 (can_group_stmts_p, vect_check_gather_scatter)
3668 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
3669 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
3670 (vect_permute_load_chain, vect_shift_permute_load_chain)
3671 (vect_transform_grouped_load)
3672 (vect_record_grouped_load_vectors): Likewise.
3673 * tree-vect-loop.c (vect_fixup_reduc_chain)
3674 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
3675 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
3676 (vectorizable_reduction, vectorizable_induction)
3677 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
3678 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
3679 (vect_get_load_store_mask): Likewise.
3680 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
3681 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
3682 * tree-vect-stmts.c (vect_mark_relevant)
3683 (is_simple_and_all_uses_invariant)
3684 (exist_non_indexing_operands_for_use_p, process_use)
3685 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
3686 (vect_get_vec_def_for_operand, vect_get_vec_defs)
3687 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
3688 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
3689 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
3690 (get_negative_load_store_type, get_load_store_type)
3691 (vect_check_load_store_mask, vect_check_store_rhs)
3692 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
3693 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
3694 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
3695 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3696 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
3697 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
3698 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
3699 (supportable_widening_operation): Likewise.
3701 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3703 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
3704 a stmt_vec_info instead of a gcall.
3705 (vect_check_gather_scatter): Update call accordingly.
3706 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
3708 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
3709 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
3710 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
3711 instead of a gimple stmt.
3712 (vect_transform_loop): Update calls accordingly.
3713 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
3714 stmt_vec_infos instead of gimple stmts.
3715 (vect_analyze_slp_instance): Update use accordingly.
3716 * tree-vect-stmts.c (read_vector_array, write_vector_array)
3717 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
3718 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
3719 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
3720 (vect_gen_widened_results_half, vect_get_loop_based_defs)
3721 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
3722 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
3723 down to subroutines.
3725 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3727 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
3728 of the worklist from a vector of gimple stmts to a vector of
3730 * tree-vect-stmts.c (vect_mark_relevant, process_use)
3731 (vect_mark_stmts_to_be_vectorized): Likewise
3733 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3735 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
3736 statement before passing it to vect_analyze_stmt.
3737 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
3738 the chain of phi vector definitions. Track the exit phi via its
3740 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
3741 STMT_VINFO_REDUC_DEF.
3742 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
3743 stmt_vec_infos to handle the statement chains.
3744 (vect_get_slp_defs): Record the first statement in the node
3745 using a stmt_vec_info.
3746 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
3747 statements here and pass their stmt_vec_info down to subroutines.
3748 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
3749 down to vect_finish_stmt_generation.
3750 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
3751 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
3752 stmt_vec_infos to subroutines.
3753 (vect_remove_stores): Use stmt_vec_infos to handle the statement
3756 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3758 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
3759 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
3760 (vect_permute_store_chain, vect_setup_realignment)
3761 (vect_permute_load_chain, vect_shift_permute_load_chain)
3762 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
3763 stmts internally, and when passing values to other vectorizer routines.
3764 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
3765 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
3766 (vect_analyze_loop_operations, get_initial_def_for_reduction)
3767 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
3768 (vectorizable_reduction, vectorizable_induction)
3769 (vectorizable_live_operation, vect_transform_loop_stmt)
3770 (vect_transform_loop): Likewise.
3771 * tree-vect-patterns.c (vect_reassociating_reduction_p)
3772 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
3773 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
3774 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3775 (vect_slp_analyze_node_operations_1): Likewise.
3776 * tree-vect-stmts.c (vect_mark_relevant, process_use)
3777 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
3778 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
3779 (vect_finish_stmt_generation_1, get_group_load_store_type)
3780 (get_load_store_type, vect_build_gather_load_calls)
3781 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
3782 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
3783 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3784 (vectorizable_store, vectorizable_load, vectorizable_condition)
3785 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
3786 (supportable_widening_operation): Likewise.
3787 (vect_get_vector_types_for_stmt): Likewise.
3788 * tree-vectorizer.h (vect_dr_behavior): Likewise.
3790 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3792 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3793 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
3794 (vect_permute_store_chain, vect_permute_load_chain)
3795 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
3796 repeated stmt_vec_info lookups.
3797 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
3798 (vect_update_ivs_after_vectorizer): Likewise.
3799 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
3800 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
3801 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
3802 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3803 (vect_bb_slp_scalar_cost): Likewise.
3804 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
3806 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3808 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
3809 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
3810 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
3811 to get gassigns and gcalls, rather than operating on generc gimple
3813 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
3814 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
3815 (vectorizable_load, vect_analyze_stmt): Likewise.
3816 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
3818 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3820 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
3821 return stmt_vec_infos rather than gimple stmts. Do not accept
3823 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
3825 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
3826 Update use of get_later_stmt.
3827 (vect_get_constant_vectors): Update call accordingly.
3828 (vect_schedule_slp_instance): Likewise
3829 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
3830 (vect_slp_analyze_instance_dependence): Likewise.
3831 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
3833 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3835 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
3836 (stmt_info_for_cost::stmt_info): ...this new field.
3837 (add_stmt_costs): Update accordingly.
3838 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
3839 (vect_get_known_peeling_cost): Likewise.
3840 (vect_estimate_min_profitable_iters): Likewise.
3841 * tree-vect-stmts.c (record_stmt_cost): Likewise.
3843 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3845 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
3846 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
3847 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
3849 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
3851 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3853 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
3854 a gimple stmt to a stmt_vec_info.
3855 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
3857 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3859 * tree-vectorizer.h (vec_info::grouped_stores): Change from
3860 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
3861 (_loop_vec_info::reduction_chains): Likewise.
3862 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
3864 * tree-vect-slp.c (vect_analyze_slp): Likewise.
3866 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3868 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
3869 a gimple stmt to a stmt_vec_info.
3870 (_stmt_vec_info::next_element): Likewise.
3871 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
3872 (vect_slp_analyze_and_verify_node_alignment)
3873 (vect_analyze_group_access_1, vect_analyze_group_access)
3874 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
3875 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
3876 (vect_supportable_dr_alignment): Update accordingly.
3877 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
3878 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
3879 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
3880 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3881 * tree-vect-slp.c (vect_build_slp_tree_1)
3882 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
3883 (vect_split_slp_store_group, vect_analyze_slp_instance)
3884 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
3885 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
3886 (get_group_load_store_type, get_load_store_type)
3887 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
3888 (vect_transform_stmt, vect_remove_stores): Likewise.
3890 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3892 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
3894 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3895 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
3896 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
3897 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
3898 (vector_alignment_reachable_p, vect_get_data_access_cost)
3899 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
3900 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
3901 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
3902 (vect_analyze_group_access_1, vect_analyze_group_access)
3903 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
3904 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
3905 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
3906 result of vect_dr_stmt and use the stmt_vec_info instead of
3907 the associated gimple stmt.
3908 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3909 (vect_gen_prolog_loop_niters): Likewise.
3910 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3912 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3914 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
3915 to a vec<stmt_vec_info>.
3916 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
3917 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
3919 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
3920 to a vec<stmt_vec_info>.
3921 (bst_traits::value_type, bst_traits::value_type): Likewise.
3922 (bst_traits::hash): Update accordingly.
3923 (vect_get_and_check_slp_defs): Change the stmts parameter from
3924 a vec<gimple *> to a vec<stmt_vec_info>.
3925 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
3926 (vect_build_slp_tree): Likewise.
3927 (vect_build_slp_tree_2): Likewise. Update uses of
3928 SLP_TREE_SCALAR_STMTS.
3929 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
3930 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
3931 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
3932 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
3933 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
3934 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
3935 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
3936 (vect_get_constant_vectors, vect_get_slp_defs)
3937 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
3938 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
3939 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
3940 instead of gimple stmts.
3941 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
3942 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
3943 (vect_slp_analyze_instance_dependence): Update uses of
3944 SLP_TREE_SCALAR_STMTS.
3945 (vect_slp_analyze_and_verify_node_alignment): Likewise.
3946 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
3947 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
3948 (get_initial_defs_for_reduction): Likewise.
3949 (vect_create_epilog_for_reduction): Likewise.
3950 (vectorize_fold_left_reduction): Likewise.
3951 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
3952 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
3953 (can_vectorize_live_stmts): Likewise.
3955 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3957 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
3958 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
3959 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
3961 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
3963 (gather_scalar_reductions): Update after above interface changes.
3964 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
3965 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
3967 (vect_force_simple_reduction): Likewise.
3968 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
3969 LOOP_VINFO_REDUCTIONS.
3970 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3972 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3974 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
3975 a gimple stmt to a stmt_vec_info.
3976 * tree-vect-loop.c (vect_active_double_reduction_p)
3977 (vect_force_simple_reduction, vectorizable_reduction): Update
3980 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3982 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
3983 vec<gimple *> to a vec<stmt_vec_info>.
3984 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
3985 the reduction_phis argument from a vec<gimple *> to a
3987 (vectorizable_reduction): Likewise the phis local variable that
3988 is passed to vect_create_epilog_for_reduction. Update for new type
3989 of SLP_TREE_VEC_STMTS.
3990 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
3991 (vectorizable_live_operation): Likewise.
3992 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
3993 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
3995 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3997 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
3998 a gimple stmt to a stmt_vec_info.
3999 (vectorizable_condition, vectorizable_live_operation)
4000 (vectorizable_reduction, vectorizable_induction): Pass back the
4001 vectorized statement as a stmt_vec_info.
4002 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
4003 use of STMT_VINFO_VEC_STMT.
4004 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
4005 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
4006 as stmt_vec_infos rather than gimple stmts.
4007 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
4009 (vectorizable_live_operation): Likewise.
4010 (vectorizable_reduction, vectorizable_induction): Likewise,
4011 updating use of STMT_VINFO_VEC_STMT.
4012 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
4013 of STMT_VINFO_VEC_STMT.
4014 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
4015 (vectorizable_simd_clone_call, vectorizable_conversion)
4016 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4017 (vectorizable_store, vectorizable_load, vectorizable_condition)
4018 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
4019 from a gimple stmt to a stmt_vec_info.
4020 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
4021 pointer to a stmt_vec_info to the vectorizable_* routines.
4023 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4025 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
4026 a gimple stmt to a stmt_vec_info.
4027 (is_pattern_stmt_p): Update accordingly.
4028 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
4029 (vect_record_grouped_load_vectors): Likewise.
4030 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
4031 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
4032 (vect_model_reduction_cost): Likewise.
4033 (vect_create_epilog_for_reduction): Likewise.
4034 (vectorizable_reduction, vectorizable_induction): Likewise.
4035 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4036 Return the stmt_vec_info for the pattern statement.
4037 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
4038 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
4039 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
4040 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
4041 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
4042 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
4043 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
4044 (free_stmt_vec_info, vect_is_simple_use): Likewise.
4046 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4048 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
4049 (vect_finish_stmt_generation): Likewise.
4050 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
4051 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
4052 (vect_build_gather_load_calls): Use the return value of the above
4053 functions instead of a separate call to vinfo_for_stmt. Use narrow
4054 scopes for the input gimple stmt and wider scopes for the associated
4055 stmt_vec_info. Use vec_info::lookup_def when setting these
4056 stmt_vec_infos from an SSA_NAME definition.
4057 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4058 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4059 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4060 (vectorizable_store, vectorizable_load, vectorizable_condition)
4061 (vectorizable_comparison): Likewise.
4062 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
4063 (vectorizable_reduction): Likewise.
4065 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4067 * tree-vectorizer.h (vect_is_simple_use): Add an optional
4068 stmt_vec_info * parameter before the optional gimple **.
4069 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4070 (process_use, vect_get_vec_def_for_operand_1): Update callers.
4071 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
4072 * tree-vect-loop.c (vectorizable_reduction): Likewise.
4073 (vectorizable_live_operation): Likewise.
4074 * tree-vect-patterns.c (type_conversion_p): Likewise.
4075 (vect_look_through_possible_promotion): Likewise.
4076 (vect_recog_rotate_pattern): Likewise.
4077 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
4079 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4081 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
4082 a typedef to a wrapper class.
4083 (NULL_STMT_VEC_INFO): New macro.
4084 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
4085 (stmt_vec_info::operator*): New function.
4086 (stmt_vec_info::operator gimple *): Likewise.
4087 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
4088 (add_stmt_costs): Likewise.
4089 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
4090 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
4091 (vect_get_known_peeling_cost): Likewise.
4092 (vect_estimate_min_profitable_iters): Likewise.
4093 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4094 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
4095 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
4096 (vectorizable_store, free_stmt_vec_infos): Likewise.
4097 (new_stmt_vec_info): Change return type of xcalloc to
4100 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4102 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
4103 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
4104 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
4105 a single_imm_use-based sequence.
4106 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
4108 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4110 * tree-vectorizer.h (vec_info::lookup_def): Declare.
4111 * tree-vectorizer.c (vec_info::lookup_def): New function.
4112 * tree-vect-patterns.c (vect_get_internal_def): Use it.
4113 (vect_widened_op_tree): Likewise.
4114 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4115 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
4116 (vectorizable_reduction): Likewise.
4117 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
4119 (vect_is_slp_reduction): Update calls accordingly. Use
4120 vec_info::lookup_def.
4121 (vect_is_simple_reduction): Likewise
4122 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
4124 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4126 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
4127 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
4128 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
4130 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
4131 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
4132 (vect_update_vf_for_slp, vect_analyze_loop_operations)
4133 (vect_is_slp_reduction, vectorizable_induction)
4134 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
4135 * tree-vect-patterns.c (vect_init_pattern_stmt):
4136 (vect_determine_min_output_precision_1, vect_determine_precisions)
4137 (vect_pattern_recog): Likewise.
4138 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
4139 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
4140 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
4141 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
4142 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
4143 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
4144 info field from a loop to a loop_vec_info.
4146 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4148 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
4149 (vec_info::add_stmt): Declare.
4150 * tree-vectorizer.c (vec_info::add_stmt): New function.
4151 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
4152 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
4153 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4154 (vectorizable_induction): Likewise.
4155 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
4156 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
4157 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
4158 (vectorizable_load): Likewise.
4159 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4160 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
4161 (vect_recog_gather_scatter_pattern): Likewise.
4162 (append_pattern_def_seq): Likewise. Remove a check that is
4163 performed by add_stmt itself.
4165 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4167 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
4168 which make_ssa_name was called with new_stmt before new_stmt
4171 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4173 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
4175 (vect_is_slp_reduction): ...here...
4176 (vect_is_simple_reduction): ...and here. Remove repetition of tests
4177 that are already known to be false.
4179 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4181 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
4182 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
4183 STMT_VINFO_NUM_SLP_USES when it's true.
4184 (vect_free_slp_instance): Add a final_p parameter and pass it to
4186 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
4187 (vect_analyze_slp_instance): Likewise.
4188 (vect_slp_analyze_operations): Likewise.
4189 (vect_slp_analyze_bb_1): Likewise.
4190 * tree-vectorizer.c (vec_info): Likewise.
4191 * tree-vect-loop.c (vect_transform_loop): Likewise.
4193 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4195 * tree-vect-loop.c (vectorizable_reduction): Assert that the
4196 function is not called for second and subsequent members of
4199 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4201 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
4202 cases for nested loops from here to ...
4203 (vect_create_epilog_for_reduction): ...here. Only call
4204 vect_is_simple_use for inner-loop reductions.
4206 2018-07-31 Martin Liska <mliska@suse.cz>
4208 PR gcov-profile/85338
4209 PR gcov-profile/85350
4210 PR gcov-profile/85372
4211 * profile.c (struct location_triplet): New.
4212 (struct location_triplet_hash): Likewise.
4213 (output_location): Do not output a BB that
4214 is already recorded for a line.
4215 (branch_prob): Use streamed_locations.
4217 2018-07-31 Martin Liska <mliska@suse.cz>
4219 PR gcov-profile/85370
4220 * coverage.c (coverage_begin_function): Do not mark target
4221 clones as artificial functions.
4223 2018-07-31 Martin Liska <mliska@suse.cz>
4225 PR gcov-profile/83813
4226 PR gcov-profile/84758
4227 PR gcov-profile/85217
4228 PR gcov-profile/85332
4229 * profile.c (branch_prob): Do not record GOTO expressions
4230 for GIMPLE statements which locations are already streamed.
4232 2018-07-31 Olivier Hainque <hainque@adacore.com>
4234 * gcc.c (handle_spec_function): Accept a soft_matched_part
4235 argument, as do_spec_1. Pass it down to ...
4236 (eval_spec_function): Accept a soft_matched_part argument,
4237 and pass it down to ...
4238 (do_spec_2): Accept a soft_matched_part argument, and pass
4239 it down to do_spec_1.
4240 (do_spec_1): Pass soft_matched_part to handle_spec_function.
4241 (handle_braces): Update call to handle_spec_function.
4242 (driver::set_up_specs): Update calls to do_spec_2.
4243 (compare_debug_dump_opt_spec_function): Likewise.
4244 (compare_debug_self_opt_spec_function): Likewise.
4246 2018-07-31 Olivier Hainque <hainque@adacore.com>
4248 * common.opt (nolibc): New option.
4249 * doc/invoke.texi (Link Options): Document it.
4250 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
4251 * config/alpha/linux.h: Likewise.
4252 * config/arc/elf.h: Likewise.
4253 * config/arm/uclinux-elf.h: Likewise.
4254 * config/arm/unknown-elf.h: Likewise.
4255 * config/avr/avrlibc.h: Likewise.
4256 * config/bfin/bfin.h: Likewise.
4257 * config/bfin/linux.h: Likewise.
4258 * config/bfin/uclinux.h: Likewise.
4259 * config/darwin.h: Likewise.
4260 * config/darwin10.h: Likewise.
4261 * config/darwin12.h: Likewise.
4262 * config/gnu-user.h: Likewise.
4263 * config/lm32/uclinux-elf.h: Likewise.
4264 * config/pa/pa-hpux11.h: Likewise.
4265 * config/pa/pa64-hpux.h: Likewise.
4266 * config/sparc/sparc.h: Likewise.
4268 2018-07-31 Olivier Hainque <hainque@adacore.com>
4270 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
4271 undefined variables.
4273 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
4276 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
4279 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
4281 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
4282 terminated string literal.
4284 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
4286 PR rtl-optimization/85160
4287 * combine.c (is_just_move): New function.
4288 (try_combine): Allow combining two instructions into two if neither of
4289 the original instructions was a move.
4291 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
4294 * doc/extend.texi (Global Register Variables): Discourage use of type
4296 (Local Register Variables): Likewise.
4298 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
4300 PR tree-optimization/86506
4301 * hwint.h (ceil_log2): Resync with hwint.c implementation.
4303 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
4306 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
4307 hard_regno, make sure no insn between `from` and `to` clobbers it.
4309 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
4310 Tom de Vries <tdevries@suse.de>
4312 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
4313 (PTX_DEFAULT_RUNTIME_DIM): ... this.
4314 (nvptx_goacc_validate_dims): Set default worker and gang dims to
4315 PTX_DEFAULT_RUNTIME_DIM.
4316 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
4318 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
4320 * config/pa/pa.c (pa_output_addr_vec): Align address table.
4321 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
4322 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
4324 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
4326 * config/rs6000/constraints.md (wG constraint): Delete, no longer
4328 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
4329 predicate to reflect toc fusion has been deleted.
4330 (toc_fusion_mem_raw): Delete, no longer used.
4331 (toc_fusion_mem_wrapped): Likewise.
4332 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
4334 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
4335 Delete, no longer used.
4336 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
4337 meant to be used for toc fusion.
4338 (rs6000_debug_print_mode): Delete toc fusion debugging.
4339 (rs6000_debug_reg_global): Likewise.
4340 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
4341 fusion and secondary reload support that were never used.
4342 (rs6000_option_override_internal): Delete TOC fusion, that was only
4343 partially defined, and it did not work unless you also used the
4345 (rs6000_legitimate_address_p): Delete TOC fusion support.
4346 (rs6000_opt_masks): Likewise.
4347 (fusion_wrap_memory_address): Delete function, no longer used.
4348 (fusion_split_address); Delete TOC fusion support.
4349 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
4350 longer used with toc fusion being deleted.
4351 (TARGET_TOC_FUSION_FP): Likewise.
4352 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
4354 (toc fusion spliter): Delete TOC fusion support.
4355 (toc_fusionload_<mode>): Likewise.
4356 (toc_fusionload_di): Likewise.
4357 (fusion_gpr_load_<mode>): Delete generator function, this insn no
4358 longer needs to be named. Rename predicate to delete TOC fusion.
4359 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
4360 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
4361 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
4362 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
4363 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
4365 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
4367 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
4368 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
4369 __int128 in built-in function prototypes.
4370 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
4371 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
4373 2018-07-27 Martin Sebor <msebor@redhat.com>
4375 PR tree-optimization/86696
4376 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
4377 types, including enums.
4378 (handle_char_store): Be prepared for the above function to fail.
4380 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
4382 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
4383 when optimization level is lower than 2 or optimize for size.
4385 2018-07-26 Martin Sebor <msebor@redhat.com>
4387 PR tree-optimization/86043
4388 PR tree-optimization/86042
4389 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
4390 (get_string_cst_length): Rename...
4391 (get_min_string_length): ...to this. Add argument.
4392 (handle_char_store): Extend to handle multi-character stores by
4394 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
4395 * tree.h (initializer_zerop): Add argument.
4397 2018-07-26 Jakub Jelinek <jakub@redhat.com>
4400 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
4401 declare target to variables if they have always,{to,from,tofrom} map
4404 2018-07-26 Martin Liska <mliska@suse.cz>
4407 * lto-wrapper.c: Add linker_output as prefix
4408 for ltrans_output_file.
4410 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
4412 PR rtl-optimization/85805
4413 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
4414 value for hard registers if that was written in the same mode.
4416 2018-07-26 Martin Liska <mliska@suse.cz>
4418 PR gcov-profile/86536
4419 * gcov.c (format_gcov): Use printf format %.*f directly
4420 and do not handle special values.
4422 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4424 * common/config/arc/arc-common.c (arc_option_optimization_table):
4425 Update default optimizations for size.
4427 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4429 * config/arc/arc.md (movsf_insn): Add short instruction selection.
4430 * config/arc/constraints.md (CfZ): New constraint.
4431 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
4432 (subsf3_fpu): Likewise.
4433 (cmpsf_fpu): Likewise.
4434 (cmpsf_fpu_uneq): Likewise.
4436 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4438 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
4440 (arc_is_uncached_mem_p): uncached applies to both the variable and
4443 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4445 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
4448 2018-07-25 David Malcolm <dmalcolm@redhat.com>
4450 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
4451 field "m_scopes" from vec to auto_vec.
4453 2018-07-25 Martin Liska <mliska@suse.cz>
4455 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
4458 2018-07-25 Richard Biener <rguenther@suse.de>
4461 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
4462 special wrt context_die late.
4463 (gen_subprogram_die): Re-use DIEs in local scope.
4465 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
4467 PR tree-optimization/86644
4468 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
4470 2018-07-25 Martin Liska <mliska@suse.cz>
4473 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
4474 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
4476 2018-07-25 Martin Liska <mliska@suse.cz>
4479 * params.def: Explain ASan abbreviation and provide
4480 a documentation link.
4482 2018-07-24 Martin Sebor <msebor@redhat.com>
4484 PR tree-optimization/86622
4485 PR tree-optimization/86532
4486 * builtins.h (string_length): Declare.
4487 * builtins.c (c_strlen): Correct handling of non-constant offsets.
4488 (check_access): Be prepared for non-constant length ranges.
4489 (string_length): Make extern.
4490 * expr.c (string_constant): Only handle the minor non-constant
4491 array index. Use string_constant to compute the length of
4492 a generic string constant.
4494 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
4496 PR tree-optimization/86618
4497 * tree-vect-stmts.c (vectorizable_call): Don't take the address
4498 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
4500 2018-07-24 David Malcolm <dmalcolm@redhat.com>
4502 PR tree-optimization/86636
4503 * json.cc (json::object::set): Fix comment. Add assertions.
4504 (json::array::append): Move here from json.h. Add comment and an
4506 (json::string::string): Likewise.
4507 * json.h (json::array::append): Move to json.cc.
4508 (json::string::string): Likewise.
4509 * optinfo-emit-json.cc
4510 (optrecord_json_writer::impl_location_to_json): Assert that we
4511 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
4512 wrapper around it. Expand the location once, rather than three
4514 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
4515 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
4517 (optrecord_json_writer::optinfo_to_json): Likewise, in four
4518 places. Fix some overlong lines.
4520 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
4522 * config/aarch64/aarch64-simd.md
4523 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
4524 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
4525 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
4526 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
4527 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
4528 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
4529 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
4530 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
4531 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
4533 2018-07-24 Jakub Jelinek <jakub@redhat.com>
4536 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
4537 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
4538 and abs_d == d, do the power of two handling if profitable.
4540 2018-07-24 Richard Biener <rguenther@suse.de>
4542 * match.pd: Add BIT_FIELD_REF canonicalizations.
4544 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
4547 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
4549 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
4551 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
4552 terminated STRING_CST object.
4554 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
4556 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
4558 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
4560 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
4561 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
4562 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
4563 the elements into a register.
4564 (rs6000_split_v4si_init_di_reg): Delete.
4565 (rs6000_split_v4si_init): Delete.
4566 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
4567 (vsx_init_v4si): Rewrite as a define_expand.
4569 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
4571 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
4572 zero_extend argument from memory): New.
4574 2018-07-22 Martin Sebor <msebor@redhat.com>
4577 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
4578 diagnosing calls with unknown arguments unless -Walloca-larger-than
4579 is restricted to less than PTRDIFF_MAX bytes.
4581 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
4583 * doc/gcov.texi (Invoking Gcov): Editorial changes.
4585 2018-07-20 David Malcolm <dmalcolm@redhat.com>
4587 * pretty-print.c (text_info::set_location): Remove redundant
4588 "line_table" parameter from call to rich_location::set_range.
4590 2018-07-20 Martin Sebor <msebor@redhat.com>
4593 * builtins.c (expand_builtin_alloca): Adjust.
4594 * calls.c (alloc_max_size): Simplify.
4595 * cgraphunit.c (cgraph_node::expand): Adjust.
4596 * common.opt (larger_than_size, warn_frame_larger_than): Remove
4598 (frame_larger_than_size): Same.
4599 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
4600 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
4602 * doc/invoke.texi (GCC Command Options): Document option arguments.
4603 Explain byte-size arguments and suffixes.
4604 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
4605 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
4606 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
4607 * doc/options.texi (UInteger): Expand.
4608 (Host_Wide_Int, ByteSize): Document new properties.
4609 * final.c (final_start_function_1): Include sizes in an error message.
4610 * function.c (frame_offset_overflow): Same.
4611 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
4612 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
4613 Diagnose unbounded alloca calls only for limits of less than
4615 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
4616 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
4618 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
4619 only for limits of less than PTRDIFF_MAX.
4620 * langhooks-def.h (lhd_handle_option): Change function argument
4622 * langhooks.c (lhd_handle_option): Same.
4623 * langhooks.h (handle_option): Same.
4624 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
4626 (var_type, var_type_struct): Same.
4627 (var_set): Handle ByteSize flag.
4628 * optc-gen.awk: Add comments to output to ease debugging. Make
4629 use of HOST_WIDE_INT where appropriate.
4630 * opts-gen-save.awk: Use %lx to format unsigned long.
4631 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
4632 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
4633 arguments. Parse bytes-size suffixes.
4634 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
4635 (enum_value_to_arg): Same.
4636 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
4637 (handle_option): Adjust.
4638 (generate_option): Change function argument to HOST_WIDE_INT.
4639 (cmdline_handle_error): Adjust.
4640 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
4641 (set_option): Change function argument to HOST_WIDE_INT.
4642 (option_enabled): Handle cl_host_wide_int.
4643 (get_option_state): Handle CLVC_SIZE.
4644 (control_warning_option): Same.
4645 * opts.c (common_handle_option): Change function argument to
4646 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
4647 OPT_Wvla_larger_than_.
4648 * opts.h (enum cl_var_type): Add an enumerator.
4649 * stor-layout.c (layout_decl): Print a more meaningful warning.
4650 * toplev.c (output_stack_usage): Adjust.
4652 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
4654 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
4655 call to inline_expand_builtin_string_cmp.
4656 (expand_builtin_strcmp): Likewise.
4657 (expand_builtin_strncmp): Likewise.
4658 (inline_string_cmp): Delete the last parameter, change char_type_node
4659 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
4661 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
4662 the inlining expansion on target where the type of the call has same or
4663 narrower precision than unsigned char.
4665 2018-07-20 David Malcolm <dmalcolm@redhat.com>
4667 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
4668 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
4669 * common.opt (fsave-optimization-record): New option.
4670 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
4671 * doc/invoke.texi (-fsave-optimization-record): New option.
4672 * dumpfile.c: Include "optinfo-emit-json.h".
4673 (struct kv_pair): Move to coretypes.h.
4674 (optgroup_options): Make non-static.
4675 (dump_context::end_scope): Call
4676 optimization_records_maybe_pop_dump_scope.
4677 * dumpfile.h (optgroup_options): New decl.
4678 * json.cc: New file.
4680 * optinfo-emit-json.cc: New file.
4681 * optinfo-emit-json.h: New file.
4682 * optinfo.cc: Include "optinfo-emit-json.h".
4683 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
4684 (optinfo_enabled_p): Check optimization_records_enabled_p.
4685 (optinfo_wants_inlining_info_p): Likewise.
4686 * optinfo.h: Update comment.
4687 * profile-count.c (profile_quality_as_string): New function.
4688 * profile-count.h (profile_quality_as_string): New decl.
4689 (profile_count::quality): New accessor.
4690 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
4691 and optinfo_emit_json_cc_tests.
4692 * selftest.h (selftest::json_cc_tests): New decl.
4693 (selftest::optinfo_emit_json_cc_tests): New decl.
4694 * toplev.c: Include "optinfo-emit-json.h".
4695 (compile_file): Call optimization_records_finish.
4696 (do_compile): Call optimization_records_start.
4697 * tree-ssa-live.c: Include optinfo.h.
4698 (remove_unused_scope_block_p): Retain inlining information if
4699 optinfo_wants_inlining_info_p returns true.
4701 2018-07-20 Richard Biener <rguenther@suse.de>
4704 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
4705 to cover -flto-partition=none.
4707 2018-07-20 Martin Liska <mliska@suse.cz>
4709 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
4710 (get_decl_source_range): Remove unused function.
4712 2018-07-20 Richard Biener <rguenther@suse.de>
4714 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
4715 (struct vn_phi_s): Likewise.
4716 (struct vn_reference_s): Likewise.
4717 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
4718 for searching the slot of an entry known to be in the hash itself.
4719 (vn_phi_hasher::equal): Likewise.
4720 (vn_reference_hasher::equal): Likewise.
4721 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
4723 (optimistic_info, current_info): Remove, keeping only valid_info.
4724 (vn_reference_lookup_1): Remove fallback lookup.
4725 (vn_reference_lookup_2): Likewise.
4726 (vn_nary_op_lookup_1): Likewise.
4727 (vn_phi_lookup): Likewise.
4728 (vn_nary_build_or_lookup_1): Make sure to not chain the built
4730 (vn_reference_insert): Adjust, chain the inserted hash element
4731 at last_inserted_ref.
4732 (vn_reference_insert_pieces): Likewise.
4733 (visit_reference_op_call): Likewise.
4734 (vn_nary_op_insert_into): Chain the inserted hash element at
4736 (vn_nary_op_insert_pieces): Adjust.
4737 (vn_nary_op_insert): Likewise.
4738 (vn_nary_op_insert_stmt): Likewise.
4739 (vn_phi_insert): Adjust, chain the inserted hash element at
4741 (process_scc): Remove clearing and copying the optimistic
4742 table. Instead remove elements inserted during an optimistic
4743 iteration from the single table we maintain.
4744 (init_scc_vn): Adjust.
4745 (free_scc_vn): Likewise.
4746 (sccvn_dom_walker::record_cond): Likewise.
4747 (sccvn_dom_walker::after_dom_children): Likewise.
4749 2018-07-19 Martin Sebor <msebor@redhat.com>
4751 PR tree-optimization/84047
4752 PR tree-optimization/83776
4753 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
4754 (check_array_bounds): Call it.
4756 2018-07-19 Martin Sebor <msebor@redhat.com>
4758 * align.h (align_flags): Use member initialization.
4760 2018-07-19 David Malcolm <dmalcolm@redhat.com>
4762 * Makefile.in (OBJS): Add optinfo.o.
4763 * coretypes.h (class symtab_node): New forward decl.
4764 (struct cgraph_node): New forward decl.
4765 (class varpool_node): New forward decl.
4766 * dump-context.h: New file.
4767 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
4769 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
4770 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
4771 (set_alt_dump_file): Likewise.
4772 (dump_context::~dump_context): New dtor.
4773 (dump_gimple_stmt): Move implementation to...
4774 (dump_context::dump_gimple_stmt): ...this new member function.
4775 Add the stmt to any pending optinfo, creating one if need be.
4776 (dump_gimple_stmt_loc): Move implementation to...
4777 (dump_context::dump_gimple_stmt_loc): ...this new member function.
4778 Start a new optinfo and add the stmt to it.
4779 (dump_gimple_expr): Move implementation to...
4780 (dump_context::dump_gimple_expr): ...this new member function.
4781 Add the stmt to any pending optinfo, creating one if need be.
4782 (dump_gimple_expr_loc): Move implementation to...
4783 (dump_context::dump_gimple_expr_loc): ...this new member function.
4784 Start a new optinfo and add the stmt to it.
4785 (dump_generic_expr): Move implementation to...
4786 (dump_context::dump_generic_expr): ...this new member function.
4787 Add the tree to any pending optinfo, creating one if need be.
4788 (dump_generic_expr_loc): Move implementation to...
4789 (dump_context::dump_generic_expr_loc): ...this new member
4790 function. Add the tree to any pending optinfo, creating one if
4792 (dump_printf): Move implementation to...
4793 (dump_context::dump_printf_va): ...this new member function. Add
4794 the text to any pending optinfo, creating one if need be.
4795 (dump_printf_loc): Move implementation to...
4796 (dump_context::dump_printf_loc_va): ...this new member function.
4797 Start a new optinfo and add the stmt to it.
4798 (dump_dec): Move implementation to...
4799 (dump_context::dump_dec): ...this new member function. Add the
4800 value to any pending optinfo, creating one if need be.
4801 (dump_context::dump_symtab_node): New member function.
4802 (dump_context::get_scope_depth): New member function.
4803 (dump_context::begin_scope): New member function.
4804 (dump_context::end_scope): New member function.
4805 (dump_context::ensure_pending_optinfo): New member function.
4806 (dump_context::begin_next_optinfo): New member function.
4807 (dump_context::end_any_optinfo): New member function.
4808 (dump_context::s_current): New global.
4809 (dump_context::s_default): New global.
4810 (dump_scope_depth): Delete global.
4811 (dumpfile_ensure_any_optinfo_are_flushed): New function.
4812 (dump_symtab_node): New function.
4813 (get_dump_scope_depth): Reimplement in terms of dump_context.
4814 (dump_begin_scope): Likewise.
4815 (dump_end_scope): Likewise.
4816 (selftest::temp_dump_context::temp_dump_context): New ctor.
4817 (selftest::temp_dump_context::~temp_dump_context): New dtor.
4818 (selftest::verify_item): New function.
4819 (ASSERT_IS_TEXT): New macro.
4820 (ASSERT_IS_TREE): New macro.
4821 (ASSERT_IS_GIMPLE): New macro.
4822 (selftest::test_capture_of_dump_calls): New test.
4823 (selftest::dumpfile_c_tests): Call it.
4824 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
4825 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
4826 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
4827 descriptive comment.
4828 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
4829 (dump_node, dump_bb): Move these unrelated decls.
4830 (class dump_manager): Add leading comment.
4831 * optinfo.cc: New file.
4832 * optinfo.h: New file.
4834 2018-07-19 Michael Collison <michael.collison@arm.com>
4835 Richard Henderson <rth@redhat.com>
4837 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
4838 (subti): Handle op1 zero.
4839 (subvti4, usub4ti4): New.
4840 (*sub<GPI>3_compare1_imm): New.
4841 (sub<GPI>3_carryinCV): New.
4842 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
4843 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
4845 2018-07-19 Michael Collison <michael.collison@arm.com>
4846 Richard Henderson <rth@redhat.com>
4848 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
4849 (addti3): Create simpler code if low part is already known to be 0.
4850 (addvti4, uaddvti4): New.
4851 (*add<GPI>3_compareC_cconly_imm): New.
4852 (*add<GPI>3_compareC_cconly): New.
4853 (*add<GPI>3_compareC_imm): New.
4854 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
4855 handle constants within this pattern..
4856 (*add<GPI>3_compareV_cconly_imm): New.
4857 (*add<GPI>3_compareV_cconly): New.
4858 (*add<GPI>3_compareV_imm): New.
4859 (add<GPI>3_compareV): New.
4860 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
4861 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
4862 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
4863 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
4864 with 'comparison' operator.
4865 (*add<GPI>3_compareV_cconly_imm): Ditto.
4866 (*add<GPI>3_compareV_cconly): Ditto.
4867 (*add<GPI>3_compareV_imm): Ditto.
4868 (add<GPI>3_compareV): Ditto.
4869 (add<mode>3_carryinC): Ditto.
4870 (*add<mode>3_carryinC_zero): Ditto.
4871 (*add<mode>3_carryinC): Ditto.
4872 (add<mode>3_carryinV): Ditto.
4873 (*add<mode>3_carryinV_zero): Ditto.
4874 (*add<mode>3_carryinV): Ditto.
4876 2018-07-19 Michael Collison <michael.collison@arm.com>
4877 Richard Henderson <rth@redhat.com>
4879 * config/aarch64/aarch64-modes.def (CC_V): New.
4880 * config/aarch64/aarch64-protos.h
4881 (aarch64_addti_scratch_regs): Declare
4882 (aarch64_subvti_scratch_regs): Declare.
4883 (aarch64_expand_subvti): Declare.
4884 (aarch64_gen_unlikely_cbranch): Declare
4885 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
4886 for signed overflow using CC_Vmode.
4887 (aarch64_get_condition_code_1): Handle CC_Vmode.
4888 (aarch64_gen_unlikely_cbranch): New function.
4889 (aarch64_addti_scratch_regs): New function.
4890 (aarch64_subvti_scratch_regs): New function.
4891 (aarch64_expand_subvti): New function.
4893 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
4895 * config/aarch64/aarch64-option-extensions.def: New entry for profile
4897 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
4898 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
4901 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
4904 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4905 address check not strict.
4907 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
4909 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
4911 (store_pair_lanes<mode>): Likewise.
4912 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
4913 enum value 'ADDR_QUERY_LDP_STP_N'.
4914 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
4915 (aarch64_print_address_internal): Add declaration.
4916 (aarch64_print_ldpstp_address): Remove.
4917 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
4918 (aarch64_print_operand): Change printing of 'y'.
4919 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
4920 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
4921 'true' rather than '1'.
4922 * config/aarch64/constraints.md (Uml): Likewise.
4923 (Uml): Rename to Umn.
4926 2018-07-19 Richard Biener <rguenther@suse.de>
4928 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
4930 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
4931 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
4932 (vn_reference_hasher): Likewise.
4933 (struct vn_tables_s): Remove obstack and alloc-pool members.
4934 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
4935 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
4936 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
4937 (vn_reference_insert_pieces): Likewise.
4938 (alloc_vn_nary_op_noinit): Adjust.
4939 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
4940 (vn_phi_eq): Adjust.
4941 (shared_lookup_phiargs): Remove.
4942 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
4943 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
4944 (visit_reference_op_call): Likewise.
4945 (copy_nary, copy_phi, copy_reference): Remove.
4946 (process_scc): Rewind the obstack when iterating. Do not
4947 copy the elements to valid_info but just move them from one
4948 hashtable to the other.
4949 (allocate_vn_table): Adjust.
4950 (free_vn_table): Likewise.
4951 (init_scc_vn): Likewise.
4952 (free_scc_vn): Likewise.
4954 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
4957 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
4958 indirect_return as function type attribute.
4959 (ix86_attribute_table): Change indirect_return to function
4961 * doc/extend.texi: Update indirect_return attribute.
4963 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
4965 * wide-int.h (widest2_int): New.
4966 * gimple-fold.c (arith_overflowed_p): Use it.
4967 * tree.h (widest2_int_cst): New.
4968 * tree-vrp.c (wide_int_binop_overflow): Rename from
4969 vrp_int_const_binop.
4970 Rewrite to work on trees.
4971 (extract_range_from_multiplicative_op_1): Abstract code to...
4972 (wide_int_range_min_max): ...here.
4973 (wide_int_range_cross_product): ...and here.
4974 (extract_range_from_binary_expr_1): Abstract overflow code to...
4975 (wide_int_range_mult_wrapping): ...here.
4976 * tree-vrp.h (wide_int_range_cross_product): New.
4977 (wide_int_range_mult_wrapping): New.
4979 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
4980 Julia Koval <julia.koval@intel.com>
4982 * config/i386/x86-tune-costs.h (skylake_memcpy,
4983 skylake_memset): Replace rep_prefix with unrolling for size 512.
4985 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
4988 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
4989 comparision with EQ_EXPR in last stmt.
4991 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
4993 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
4994 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
4995 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
4996 previously known as "PowerPC AltiVec Built-in Functions". Move
4997 some material to new subsubsections "PowerPC AltiVec Built-in
4998 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
5000 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
5001 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
5002 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
5003 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
5005 2018-07-18 Richard Biener <rguenther@suse.de>
5007 PR tree-optimization/86557
5008 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
5011 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
5013 * config/s390/s390.c (s390_function_profiler): Generate CFI.
5015 2018-07-17 Jeff Law <law@redhat.com>
5017 * config/arm/arm.c (get_label_padding): Update for recent
5018 changes to label_to_alignment.
5020 PR tree-optimization/86010
5021 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
5023 * config/mips/mips.c (vr4130_align_insns): Update for recent
5024 changes to label_to_alignment.
5026 * config/frv/frv.c (frv_label_align): Update for recent changes
5027 to label_to_alignment.
5029 * config/nios2/nios2.c (nios2_label_align): Update for recent
5030 changes which dropped ALIGN_LABELS_LOG.
5032 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
5034 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
5035 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
5037 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
5039 * config/arc/arc.c (arc_label_align): Use align_labels instead of
5040 deprecated align_labels_log.
5042 2018-07-17 Richard Biener <rguenther@suse.de>
5045 * dwarf2out.c (init_sections_and_labels): Always generate
5046 a debug_line_str_section for early LTO debug.
5047 (dwarf2out_finish): Reset debug_line_str_hash output early.
5048 Bump counter for extra dwarf5 .debug_loc labels to not conflict
5050 (dwarf2out_early_finish): Output debug_line_str.
5052 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
5054 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
5055 index register on z196 or later.
5057 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
5059 * config/s390/s390.c (s390_default_align): Set default function
5061 (s390_override_options_after_change): Call s390_default align.
5062 (s390_option_override_internal): Call s390_default_align.
5063 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
5065 2018-07-17 Jakub Jelinek <jakub@redhat.com>
5068 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
5069 to _looptemp_ clauses, other than the first two.
5071 2018-07-17 Martin Liska <mliska@suse.cz>
5073 * opts.c: Do not enable OPT_falign_* for -Os.
5075 2018-07-17 Martin Liska <mliska@suse.cz>
5077 * align.h (MAX_CODE_ALIGN): New.
5078 (MAX_CODE_ALIGN_VALUE): New.
5079 * common/config/i386/i386-common.c (ix86_handle_option):
5080 (MAX_CODE_ALIGN): Moved to align.h.
5081 * final.c (MAX_CODE_ALIGN): Likewise.
5082 * opts.c (parse_and_check_align_values):
5083 (MAX_CODE_ALIGN): Likewise.
5084 (MAX_CODE_ALIGN_VALUE): Likewise.
5086 2018-07-17 Martin Liska <mliska@suse.cz>
5088 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
5089 in order to fulfil coding style.
5090 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
5091 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5092 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
5093 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
5094 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
5095 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
5096 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
5097 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
5098 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5100 2018-07-17 Martin Liska <mliska@suse.cz>
5102 * align.h: New file.
5103 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
5105 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
5106 align_flags of label_to_alignment.
5107 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
5109 * config/m68k/m68k.c: Do not use removed align_labels_value and
5111 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
5112 (LOOP_ALIGN): Likewise.
5113 (LABEL_ALIGN): Likewise.
5114 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
5115 Remove not used macro.
5116 (rs6000_loop_align): Change return type to align_flags.
5117 (rs6000_loop_align_max_skip): Remove.
5118 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
5119 Change return type to align_flags.
5120 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
5121 Remove not used macro.
5122 (rs6000_loop_align): Change return type to align_flags.
5123 (rs6000_loop_align_max_skip): Remove.
5124 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
5125 * config/rx/rx-protos.h (rx_align_for_label): Make it
5127 * config/rx/rx.c (rx_align_for_label): Change return type
5129 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
5131 into align_flags class.
5132 (LABEL_ALIGN): Likewise.
5133 (LOOP_ALIGN): Likewise.
5134 * config/s390/s390.c (s390_label_align): Use align_flags
5136 (s390_asm_output_function_label): Likewise.
5137 * config/sh/sh.c (sh_override_options_after_change):
5138 Use align_flags class directly without macros.
5139 (find_barrier): Likewise.
5140 (barrier_align): Likewise.
5141 (sh_loop_align): Likewise.
5142 * config/spu/spu.c (spu_option_override):
5143 Use align_flags_tuple::get_value instead of removed macros.
5144 (spu_sched_init): Likewise.
5145 * config/spu/spu.h (GTY): Likewise.
5146 * config/visium/visium.c (visium_option_override):
5147 Set "8" as default secondary alignment.
5148 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
5149 in order to guarantee secondary alignment of 8.
5150 * coretypes.h: Include align.h header file.
5151 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
5152 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
5153 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
5154 * doc/tm.texi.in: Likewise.
5155 * final.c (struct label_alignment): Remove not used structure.
5156 (LABEL_ALIGN): Change type to align_flags.
5157 (LOOP_ALIGN): Likewise.
5158 (JUMP_ALIGN): Likewise.
5159 (default_loop_align_max_skip): Remove.
5160 (default_label_align_max_skip): Likewise.
5161 (default_jump_align_max_skip): Likewise.
5162 (default_label_align_after_barrier_max_skip):
5163 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
5164 (LABEL_TO_MAX_SKIP): Remove.
5165 (label_to_alignment): Return align_flags type instead of integer.
5166 (label_to_max_skip): Remove.
5167 (align_fuzz): Use align_flags type.
5168 (compute_alignments): Use align_flags type and use align_flags::max
5169 to combine multiple alignments.
5170 (grow_label_align): Grow vec instead of C array.
5171 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
5172 (shorten_branches): Use align_flags type and use align_flags::max
5173 to combine multiple alignments.
5174 (final_scan_insn_1): Remove usage of secondary alignment that comes
5175 from label alignment, but instead use proper secondary alignment
5176 which is computed in grow_label_align.
5177 * flags.h (struct align_flags_tuple): Move to align.h.
5178 (struct align_flags): Likewise.
5179 (state_align_loops): Rename to align_loops.
5180 (state_align_jumps): Rename to align_jumps.
5181 (state_align_labels): Rename to align_labels.
5182 (state_align_functions): Rename to align_functions.
5183 (align_loops_log): Remove.
5184 (align_jumps_log): Remove.
5185 (align_labels_log): Remove.
5186 (align_functions_log): Remove.
5187 (align_loops_max_skip): Remove.
5188 (align_jumps_max_skip): Remove.
5189 (align_labels_max_skip): Remove.
5190 (align_functions_max_skip): Remove.
5191 (align_loops_value): Remove.
5192 (align_jumps_value): Remove.
5193 (align_labels_value): Remove.
5194 (align_functions_value): Remove.
5195 * output.h (label_to_alignment): Change return type to align_flags.
5196 (label_to_max_skip): Remove.
5197 * target.def: Remove loop_align_max_skip, label_align_max_skip,
5198 jump_align_max_skip macros.
5199 * targhooks.h (default_loop_align_max_skip): Remove.
5200 (default_label_align_max_skip): Likewise.
5201 (default_jump_align_max_skip): Likewise.
5202 (default_label_align_after_barrier_max_skip): Remove.
5203 * toplev.c (read_log_maxskip): Use ::normalize function.
5204 (parse_N_M): Remove not used argument and also call ::normalize.
5205 (parse_alignment_opts): Do not pass unused arguments.
5206 * varasm.c (assemble_start_function): Use directly align_functions
5207 instead of removed macros.
5208 * system.h: Do not poison removed macros.
5210 2018-07-17 Jakub Jelinek <jakub@redhat.com>
5213 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
5214 and cond temporaries don't have reference type if iterator has
5215 pointer type. For init use &for_pre_body instead of pre_p if
5216 for_pre_body is non-empty.
5218 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5220 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
5221 double-double modes to SFmode directly directly.
5222 (trunc<mode>sf2_fprs): Delete.
5224 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5226 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
5227 for conversions between IFmode and the decimal floating point modes.
5228 (init_float128_ieee): Use the correct names for conversions between
5229 KFmode and the decimal floating point modes.
5231 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5233 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
5234 for the conversions between TDmode and IFmode.
5235 (init_float128_ieee): Use more correct names for the conversions
5236 between TDmode and KFmode.
5238 2018-07-16 Jakub Jelinek <jakub@redhat.com>
5240 PR tree-optimization/86526
5241 * builtins.c (expand_builtin_memcmp): Formatting fixes.
5242 (inline_expand_builtin_string_cmp): Likewise.
5243 (inline_string_cmp): Likewise. Use c_readstr instead of
5244 builtin_memcpy_read_str. Add unit_mode temporary.
5246 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
5249 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
5250 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
5252 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5254 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
5255 Alphabetize prototypes of built-in functions, separating out
5256 built-in functions that are listed in this section but should be
5257 described elsewhere.
5259 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
5262 * expmed.c (emit_store_flag): Do not emit setcc followed by a
5263 conditional move when trapping comparison was split to a
5264 non-trapping one (and vice versa).
5266 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5268 * config/s390/s390.c (s390_function_profiler): Generate nops
5269 instead of profiler call sequences.
5270 * config/s390/s390.opt: Add the new option.
5272 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5274 * config/s390/s390.c (s390_function_profiler): Generate
5275 __mcount_loc section.
5276 * config/s390/s390.opt: Add the new option.
5278 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5280 * common.opt: Add the new warning.
5281 * config/s390/s390.c (s390_function_profiler): Emit "brasl
5282 %r0,__fentry__" when -mfentry is specified.
5283 (s390_option_override_internal): Disallow -mfentry for 31-bit
5285 * config/s390/s390.opt: Add the new option.
5287 2018-07-16 Richard Biener <rguenther@suse.de>
5290 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
5291 for function-local FUNCTION_DECL and RESULT_DECL immediately.
5293 2018-07-16 Martin Liska <mliska@suse.cz>
5296 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
5299 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
5301 * config/arc/arcHS.md: Update ARCHS scheduling rules.
5303 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
5305 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
5307 * config/arc/arc-cpus.def (hs4x): New cpu.
5309 * config/arc/arc-tables.opt: Regenerate.
5310 * config/arc/arc.c (arc_sched_issue_rate): New function.
5311 (TARGET_SCHED_ISSUE_RATE): Define.
5312 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
5313 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
5315 (attr tune): Add ARCHS4x tune values.
5316 (attr tune_dspmpy): Define.
5317 (*tst): Correct instruction type.
5318 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
5319 * config/arc/arcHS4x.md: New file.
5320 * config/arc/fpu.md: Update instruction type attributes.
5321 * config/arc/t-multilib: Regenerate.
5323 2018-07-16 Tom de Vries <tdevries@suse.de>
5326 * var-tracking.c (vt_initialize): Fix pre_dec handling.
5328 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5330 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
5333 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
5335 PR tree-optimization/86514
5336 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
5337 conversion to a boolean type from a type with greater precision.
5339 2018-07-16 Tom de Vries <tdevries@suse.de>
5341 * var-tracking.c (vt_initialize): Print adjusted insn slim if
5342 dump_flags request TDF_SLIM.
5344 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
5346 * fold-const.c (int_const_binop_1): Abstract...
5347 (wide_int_binop): ...wide int code here.
5348 (poly_int_binop): ...poly int code here.
5349 Abstract the rest of int_const_binop_1 into int_const_binop.
5350 * fold-const.h (wide_int_binop): New.
5351 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
5352 Remove useless PLUS/MINUS_EXPR case.
5353 (zero_nonzero_bits_from_vr): Move wide int code...
5354 (zero_nonzero_bits_from_bounds): ...here.
5355 (extract_range_from_binary_expr_1): Move mask optimization code...
5356 (range_easy_mask_min_max): ...here.
5357 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
5358 (range_easy_mask_min_max): New.
5360 2018-07-15 Jeff Law <law@redhat.com>
5363 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
5366 2018-07-14 Jim Wilson <jimw@sifive.com>
5368 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
5370 2018-07-14 Paul Koning <ni1d@arrl.net>
5372 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
5374 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
5376 * lto-streamer-out.c (copy_function_or_variable): Dump info about
5379 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
5380 Steve Munroe <munroesj52@gmail.com>
5382 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
5383 (_mm_andnot_si128): Likewise.
5384 (_mm_or_si128): Likewise.
5385 (_mm_xor_si128): Likewise.
5387 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
5390 * builtins.c (expand_builtin_memcmp): Inline the calls first
5391 when result_eq is false.
5392 (expand_builtin_strcmp): Inline the calls first.
5393 (expand_builtin_strncmp): Likewise.
5394 (inline_string_cmp): New routine. Expand a string compare
5395 call by using a sequence of char comparison.
5396 (inline_expand_builtin_string_cmp): New routine. Inline expansion
5397 a call to str(n)cmp/memcmp.
5398 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
5400 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
5402 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
5404 * config/arm/driver-arm.c: Include arm-native.h.
5405 (host_detect_local_cpu): Use auto-generated data tables.
5406 (vendors, arm_cpu_table): Delete. Move part information to ...
5407 * config/arm/arm-cpus.in: ... here.
5408 * config/arm/parsecpu.awk (gen_native): New function.
5409 (vendor, part): New CPU fields.
5410 (END): Add support for building the native CPU detection tables.
5411 * config/arm/t-arm (arm-native.h): Add build rule.
5412 (driver-arm.o): Add dependency on arm-native.h.
5414 2018-07-13 Richard Biener <rguenther@suse.de>
5417 * match.pd (addr1 - addr2): Allow either of the operand to
5420 2018-07-13 Tom de Vries <tdevries@suse.de>
5422 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
5425 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
5427 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
5428 arrays instead of numbered variables.
5430 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
5432 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
5433 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
5435 2018-07-13 Richard Biener <rguenther@suse.de>
5438 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
5439 instead of get_context_die.
5441 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
5442 Richard Biener <rguenther@suse.de>
5445 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
5446 that the loop latch destination where phi is defined.
5448 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
5450 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
5451 (riscv_expand_epilogue): Add assertion to check interrupt mode.
5452 (riscv_set_current_function): Extract getting interrupt type to new
5454 (riscv_get_interrupt_type): New function.
5455 (riscv_merge_decl_attributes): New function, checking interrupt type is
5457 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
5459 2018-07-12 Paul Koning <ni1d@arrl.net>
5461 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
5464 2018-07-12 Paul Koning <ni1d@arrl.net>
5466 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
5467 zero reference, add doloop_end instead.
5468 * doc/md.texi (decrement_and_branch_until_zero): Remove.
5469 (Looping patterns): Remove decrement_and_branch_until_zero. Add
5470 detail for doloop_end.
5472 2018-07-12 Martin Sebor <msebor@redhat.com>
5475 * attribs.c (decl_attributes): Reject conflicting attributes before
5476 calling attribute handlers.
5478 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
5480 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
5482 (gcc::dump_manager::get_dump_file_name): likewise.
5483 (dump_begin): Likewise.
5484 * dumpfile.h (dump_begin): Update prototype.
5485 (gcc::dump_manager::get_dump_file_name,
5486 gcc::dump_manager::get_dump_file_name): Update prototype.
5488 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5490 * internal-fn.h (vectorizable_internal_fn_p): New function.
5491 * tree-vect-slp.c (compatible_calls_p): Likewise.
5492 (vect_build_slp_tree_1): Remove nops argument. Handle calls
5493 to internal functions.
5494 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
5496 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5498 * fold-const.h (inverse_conditions_p): Declare.
5499 * fold-const.c (inverse_conditions_p): New function.
5500 * match.pd: Use inverse_conditions_p. Add folds of view_converts
5501 that test the inverse condition of a conditional internal function.
5502 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
5503 * internal-fn.c (internal_fn_mask_index): Handle conditional
5505 (vectorized_internal_fn_supported_p): New function.
5506 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
5507 (any_pred_load_store): Replace with...
5508 (need_to_predicate): ...this new variable.
5509 (redundant_ssa_names): New variable.
5510 (ifcvt_can_use_mask_load_store): Move initial checks to...
5511 (ifcvt_can_predicate): ...this new function. Handle tree codes
5512 for which a conditional internal function exists.
5513 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
5514 instead of ifcvt_can_use_mask_load_store. Update after variable
5516 (predicate_load_or_store): New function, split out from
5517 predicate_mem_writes.
5518 (check_redundant_cond_expr): New function.
5519 (value_available_p): Likewise.
5520 (predicate_rhs_code): Likewise.
5521 (predicate_mem_writes): Rename to...
5522 (predicate_statements): ...this. Use predicate_load_or_store
5523 and predicate_rhs_code.
5524 (combine_blocks, tree_if_conversion): Update after above name changes.
5525 (ifcvt_local_dce): Handle redundant_ssa_names.
5526 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
5527 general conditional functions.
5528 * tree-vect-stmts.c (vectorizable_call): Likewise.
5530 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5531 Alan Hayward <alan.hayward@arm.com>
5532 David Sherwood <david.sherwood@arm.com>
5534 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
5535 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
5536 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
5537 plus and minus and convert them into IFN_COND_FMA-based sequences.
5538 (convert_mult_to_fma): Handle conditional plus and minus.
5540 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5542 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
5543 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
5544 (cond_fnms_optab): New optabs.
5545 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
5547 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
5548 * internal-fn.h (get_conditional_internal_fn): Declare.
5549 (get_unconditional_internal_fn): Likewise.
5550 * internal-fn.c (cond_ternary_direct): New macro.
5551 (expand_cond_ternary_optab_fn): Likewise.
5552 (direct_cond_ternary_optab_supported_p): Likewise.
5553 (FOR_EACH_COND_FN_PAIR): Likewise.
5554 (get_conditional_internal_fn): New function.
5555 (get_unconditional_internal_fn): Likewise.
5556 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
5557 (gimple_match_op::gimple_match_op): Add a new overload for 5
5559 (gimple_match_op::set_op): Likewise.
5560 (gimple_resimplify5): Declare.
5561 * genmatch.c (decision_tree::gen): Generate simplifications for
5563 * gimple-match-head.c (gimple_simplify): Define an overload for
5564 5 operands. Handle calls with 5 arguments in the top-level overload.
5565 (convert_conditional_op): Handle conversions from unconditional
5566 internal functions to conditional ones.
5567 (gimple_resimplify5): New function.
5568 (build_call_internal): Pass a fifth operand.
5569 (maybe_push_res_to_seq): Likewise.
5570 (try_conditional_simplification): Try converting conditional
5571 internal functions to unconditional internal functions.
5572 Handle 3-operand unconditional forms.
5573 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
5574 Define ternary equivalents of the current rules for binary conditional
5576 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
5578 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
5579 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
5580 (optab): Handle them.
5581 (SVE_COND_FP_TERNARY): New int iterator.
5582 (sve_fmla_op, sve_fmad_op): New int attributes.
5583 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
5584 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
5585 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
5587 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5589 * target.def (preferred_else_value): New target hook.
5590 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
5591 * doc/tm.texi: Regenerate.
5592 * targhooks.h (default_preferred_else_value): Declare.
5593 * targhooks.c (default_preferred_else_value): New function.
5594 * internal-fn.h (conditional_internal_fn_code): Declare.
5595 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
5596 (get_conditional_internal_fn): Use it.
5597 (conditional_internal_fn_code): New function.
5598 * gimple-match.h (gimple_match_cond): New struct.
5599 (gimple_match_op): Add a cond member function.
5600 (gimple_match_op::gimple_match_op): Update all forms to take a
5602 * genmatch.c (expr::gen_transform): Use the same condition as res_op
5603 for the suboperation, but don't specify a particular else_value.
5604 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
5605 (visit_nary_op, visit_reference_op_load): Pass
5606 gimple_match_cond::UNCOND to the gimple_match_op constructor.
5607 * gimple-match-head.c: Include tree-eh.h
5608 (convert_conditional_op): New function.
5609 (maybe_resimplify_conditional_op): Likewise.
5610 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
5611 (gimple_resimplify2): Likewise.
5612 (gimple_resimplify3): Likewise.
5613 (gimple_resimplify4): Likewise.
5614 (maybe_push_res_to_seq): Return null for conditional operations.
5615 (try_conditional_simplification): New function.
5616 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
5618 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
5620 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
5622 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
5624 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
5626 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
5628 (hash_tree): Do not hash DECL_FCONTEXT
5629 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
5630 Do not stream DECL_FCONTEXT.
5631 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
5632 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
5634 2018-07-12 Richard Biener <rguenther@suse.de>
5637 * dwarf2out.c (gen_block_die): Only output blocks when they have
5638 at least one !DECL_IGNORED_P variable.
5640 2018-07-12 Richard Biener <rguenther@suse.de>
5643 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
5644 Remove -mieee-fp handling.
5646 2018-07-12 Richard Biener <rguenther@suse.de>
5648 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
5649 left-over from last patch.
5651 2018-07-12 Jakub Jelinek <jakub@redhat.com>
5653 PR tree-optimization/86492
5654 * gimple-ssa-store-merging.c
5655 (imm_store_chain_info::coalesce_immediate_stores): Call
5656 check_no_overlap even for the merge_overlapping case. Formatting fix.
5658 2018-07-12 Richard Biener <rguenther@suse.de>
5661 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
5662 move possibly trapping operations into the conditional.
5664 2018-07-12 Richard Biener <rguenther@suse.de>
5666 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
5667 (vn_lookup_simplify_result): Remove recursion limit applied
5669 (vn_nary_build_or_lookup_1): Adjust.
5670 (try_to_simplify): Likewise.
5671 * gimple-match-head.c (gimple_resimplify1): Instead apply one
5673 (gimple_resimplify2): Likewise.
5674 (gimple_resimplify3): Likewise.
5675 (gimple_resimplify4): Likewise.
5677 2018-07-11 Jakub Jelinek <jakub@redhat.com>
5679 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
5680 Use __mmask64 type instead of __mmask8 for __M argument.
5681 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
5682 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
5684 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
5685 __mmask16 for __M argument.
5686 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
5687 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
5688 to __mmask16 instead of __mmask8.
5689 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
5690 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
5691 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
5692 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
5693 instead of __mmask16 for __U argument.
5694 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
5695 __mmask16 instead of __mmask8 for __U argument.
5696 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
5698 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
5700 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
5702 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
5704 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
5706 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
5708 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
5710 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
5712 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
5714 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
5716 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
5717 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
5718 return type as well as __M argument type and all casts from __mmask8
5720 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
5721 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
5722 return type as well as __M argument type and all casts from __mmask8
5724 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
5725 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
5726 return type as well as __M argument type and all casts from __mmask8
5728 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
5729 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
5730 return type as well as __M argument type and all casts from __mmask8
5732 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
5733 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
5736 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
5738 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
5739 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
5742 2018-07-11 Paul Koning <ni1d@arrl.net>
5744 * doc/md.texi (define_subst): Document how multiple occurrences of
5745 the same argument in the replacement pattern are handled.
5747 2018-07-11 Paul Koning <ni1d@arrl.net>
5749 * doc/extend.texi (Common Variable Attributes): Move "mode" into
5751 (Common Type Attributes): Add "mode" attribute.
5753 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
5755 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
5756 stream DECL_ORIGINAL_TYPE.
5757 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
5758 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
5759 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
5760 Do not walk original type.
5761 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
5763 (write_ts_decl_non_common_tree_pointers): Do not stream
5765 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
5766 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
5768 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
5770 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
5771 thread twice from the same starting edge.
5773 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
5775 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
5776 * gimple.c (gimple_call_nonnull_result_p): ...here...
5777 (gimple_call_nonnull_arg): ...and here.
5778 * gimple.h (gimple_call_nonnull_result_p): New.
5779 (gimple_call_nonnull_arg): New.
5781 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
5783 * config/arm/arm-cpus.in: Move information from fpu field of each
5784 cpu definition to the isa field.
5785 * config/arm/parsecpu.awk (fpu): Delete match rule.
5786 (gen_comm_data): Don't add bits from the CPU's FPU entry.
5788 2018-07-11 Richard Biener <rguenther@suse.de>
5791 * dwarf2out.c (init_sections_and_labels): Use
5792 output_asm_line_debug_info consistently.
5793 (dwarf2out_early_finish): Likewise.
5794 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
5797 2018-07-11 Richard Biener <rguenther@suse.de>
5799 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
5800 Rework father_bb setting in a way to avoid propagating constants
5801 multiple times on a loop body.
5803 2018-07-10 Mark Wielaard <mark@klomp.org>
5806 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
5808 2018-07-10 Richard Biener <rguenther@suse.de>
5810 * hash-map.h (hash_map::iterator::operator*): Return
5811 references to key and value.
5813 2018-07-10 Jakub Jelinek <jakub@redhat.com>
5816 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
5817 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
5818 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
5819 (gimplify_omp_for): For composite loops, move outer
5820 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
5821 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
5822 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
5823 TREE_LIST for both the original class iterator and the "last" helper
5824 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
5825 loop, remember has_decl_expr from outer composite loops for the
5826 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
5828 2018-07-09 Martin Sebor <msebor@redhat.com>
5832 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
5833 accessing implicitly initialized array elements.
5834 * expr.c (string_constant): Handle string initializers of
5835 character arrays within aggregates.
5836 * gimple-fold.c (fold_array_ctor_reference): Add argument.
5837 Store element offset. As a special case, handle zero size.
5838 (fold_nonarray_ctor_reference): Same.
5839 (fold_ctor_reference): Add argument. Store subobject offset.
5840 * gimple-fold.h (fold_ctor_reference): Add argument.
5842 2018-07-09 Paul Koning <ni1d@arrl.net>
5844 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
5845 (pdp11_insn_cost): New function.
5846 (pdp11_md_asm_adjust): New function.
5847 (TARGET_INVALID_WITHIN_DOLOOP): Define.
5848 (pdp11_rtx_costs): Update to match machine better.
5849 (output_addr_const_pdp11): Correct format mismatch warnings.
5850 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
5851 * config/pdp11/pdp11.md: General change to add base_cost and/or
5852 length attributes for use by new pdp11_insn_cost function.
5853 (MIN_BRANCH): Correct definition.
5855 (doloop_end): Use standard pattern name for looping pattern.
5856 (doloop_end_nocc): New.
5857 (movsf): Add another constraint alternative.
5858 (zero_extendqihi2): Add constraint alternatives for not in place
5860 (zero_extendhisi2): Remove.
5861 (shift patterns): Add CC handling variants.
5865 (define_peephole2): New peephole to recognize mov that sets CC for
5868 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5870 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
5871 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
5874 2018-07-09 Jakub Jelinek <jakub@redhat.com>
5877 * real.c (real_nextafter): Return true if result is denormal.
5879 2018-07-09 Martin Liska <mliska@suse.cz>
5881 * common.opt: Add back wrongly removed attribute.
5883 2018-07-09 Richard Biener <rguenther@suse.de>
5886 * dwarf2out.c (gen_block_die): For an early generated DIE
5887 always output high/low PC attributes.
5889 2018-07-09 Tom de Vries <tdevries@suse.de>
5891 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
5892 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
5893 onto VAR_DECL with abstract origin.
5895 2018-07-07 Jim Wilson <jimw@sifive.com>
5897 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
5899 2018-07-07 Tom de Vries <tdevries@suse.de>
5901 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
5903 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
5905 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
5907 (combine_bound): Use wide-int overflow calculation instead of
5909 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
5911 * fold-const.c (int_const_binop_2): Same.
5912 (extract_muldiv_1): Same.
5913 (fold_div_compare): Same.
5914 (fold_abs_const): Same.
5916 * poly-int.h (add): Same.
5920 * predict.c (predict_iv_comparison): Same.
5921 * profile-count.c (slow_safe_scale_64bit): Same.
5922 * simplify-rtx.c (simplify_const_binary_operation): Same.
5923 * tree-chrec.c (tree_fold_binomial): Same.
5924 * tree-data-ref.c (split_constant_offset_1): Same.
5925 * tree-if-conv.c (idx_within_array_bound): Same.
5926 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
5927 * tree-ssa-phiopt.c (minmax_replacement): Same.
5928 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
5929 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
5930 * vr-values.c (vr_values::adjust_range_with_scev): Same.
5931 * wide-int.cc (wi::add_large): Same.
5932 (wi::mul_internal): Same.
5933 (wi::sub_large): Same.
5934 (wi::divmod_internal): Same.
5935 * wide-int.h: Change overflow type to overflow_type for neg, add,
5936 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
5937 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
5938 mul_internal, divmod_internal.
5939 (overflow_type): New enum.
5940 (accumulate_overflow): New.
5942 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
5944 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
5945 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
5947 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
5949 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
5950 argument is checked for zero before entering loop, avoid checking again.
5952 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
5954 * gimplify.h (generic_expr_could_trap_p): Set as global function.
5955 * gimplify.h (generic_expr_could_trap_p): Likwise.
5956 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
5958 2018-07-06 Jakub Jelinek <jakub@redhat.com>
5960 PR tree-optimization/86401
5961 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
5962 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
5963 (fold_bit_and_mask): ... here. New helper function for match.pd.
5964 * fold-const.h (fold_bit_and_mask): Declare.
5965 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
5967 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
5970 * target.def (translate_mode_attribute): New hook.
5971 * targhooks.h (default_translate_mode_attribute): Declare.
5972 * targhooks.c (default_translate_mode_attribute): New function.
5973 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
5974 * doc/tm.texi: Regenerate.
5975 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
5976 (rs6000_translate_mode_attribute): New function.
5978 2018-07-06 Paul Koning <ni1d@arrl.net>
5980 * doc/md.texi (define_split): Document DONE and FAIL.
5981 (define_peephole2): Ditto.
5983 2018-07-05 Jeff Law <law@redhat.com>
5985 PR tree-optimization/86010
5986 * tree-ssa-dse.c (compute_trims): More aggressively trim at
5987 both the head and tail of mem* and str* calls.
5989 2018-07-05 Jim Wilson <jimw@sifive.com>
5991 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
5993 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
5995 * config/aarch64/aarch64-simd.md: correct flags text for
5996 MIN_EXPR replacement.
5998 2018-07-05 James Clarke <jrtc27@jrtc27.com>
6000 * configure: Regenerated.
6002 2018-07-05 Carl Love <cel@us.ibm.com>
6004 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
6005 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
6006 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
6007 VSX_BUILTIN_DOUBLEL_V4SF.
6009 2018-07-05 Martin Sebor <msebor@redhat.com>
6012 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
6013 than its domain to compute its the upper bound of a char array.
6015 2018-07-05 Nathan Sidwell <nathan@acm.org>
6017 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
6018 * doc/cpp.texi: Update comment.
6019 * doc/tm.texi: Rebuilt.
6020 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
6021 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
6022 * doc/extend.texi (Backwards Compatibility): Clarify it is system
6023 headers affected by extern "C".
6024 * system.h: Poison NO_IMPLICIT_EXTERN_C.
6025 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
6026 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
6027 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
6028 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
6029 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
6030 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
6031 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
6032 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
6033 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
6034 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
6035 config/sparc/sp64-elf.h, config/spu/spu.h,
6036 config/stormy16/stormy16.h, config/v850/v850.h,
6037 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
6038 define NO_IMPLICIT_EXTERN_C.
6039 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
6041 2018-07-05 Tamar Christina <tamar.christina@arm.com>
6044 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6045 instead of GET_MODE_SIZE when comparing Units.
6047 2018-07-05 Tamar Christina <tamar.christina@arm.com>
6050 * rtlanal.c (set_noop_p): Constrain on mode change,
6051 include hard-reg-set.h
6053 2018-07-05 Tamar Christina <tamar.christina@arm.com>
6055 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
6057 2018-07-05 Jakub Jelinek <jakub@redhat.com>
6060 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
6063 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
6065 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
6067 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
6070 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
6072 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
6074 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
6077 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
6079 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
6081 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
6082 neg_*_op* variables.
6084 2018-07-04 Martin Liska <mliska@suse.cz>
6086 * tree-switch-conversion.c: Define
6087 max_ratio_for_speed and max_ratio_for_size constants.
6089 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
6090 Martin Liska <mliska@suse.cz>
6095 * common.opt: Rename align options with 'str_' prefix.
6096 * common/config/i386/i386-common.c (set_malign_value): New
6098 (ix86_handle_option): Use it to set -falign-* options/
6099 * config/aarch64/aarch64-protos.h (struct tune_params): Change
6100 type from int to string.
6101 * config/aarch64/aarch64.c: Update default values from int
6103 * config/alpha/alpha.c (alpha_override_options_after_change):
6105 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
6106 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6107 max skip conditionally.
6108 * config/i386/freebsd.h (SUBALIGN_LOG): New.
6109 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6110 max skip conditionally.
6111 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6112 max skip conditionally.
6113 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
6114 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6115 max skip conditionally.
6116 * config/i386/i386.c (struct ptt): Change type from int to
6118 (ix86_default_align): Set default values.
6119 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
6120 max skip conditionally.
6121 * config/i386/iamcu.h (SUBALIGN_LOG): New.
6122 (ASM_OUTPUT_MAX_SKIP_ALIGN):
6123 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
6124 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6125 max skip conditionally.
6126 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
6127 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
6128 * config/i386/x86-64.h (SUBALIGN_LOG): New.
6129 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6130 max skip conditionally.
6131 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6132 * config/ia64/ia64.c (ia64_option_override): Set default values
6133 for alignment options.
6134 * config/m68k/m68k.c: Handle new str_align_* options.
6135 * config/mips/mips.c (mips_set_compression_mode): Change
6137 (mips_option_override): Set default values for options.
6138 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
6140 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6142 * config/rx/rx.c (rx_option_override): Likewise.
6143 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
6144 (LABEL_ALIGN): Use align_labels_log.
6145 (LOOP_ALIGN): Use align_loops_align.
6146 * config/s390/s390.c (s390_asm_output_function_label): Use new
6148 * config/sh/sh.c (sh_override_options_after_change):
6149 Change type of constants.
6150 * config/spu/spu.c (spu_sched_init): Likewise.
6151 * config/sparc/sparc.c (sparc_option_override): Set default
6153 * config/visium/visium.c (visium_option_override): Likewise.
6154 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
6155 emit p2align format with last argument if it's not needed.
6156 * doc/invoke.texi: Document extended format of -falign-*.
6157 * final.c: Use align_labels alignment.
6158 * flags.h (struct target_flag_state): Change type to use
6160 (struct align_flags_tuple): New.
6161 (struct align_flags): Likewise.
6162 (align_loops_log): Redefine macro to use new types.
6163 (align_loops_max_skip): Redefine macro to use new types.
6164 (align_jumps_log): Redefine macro to use new types.
6165 (align_jumps_max_skip): Redefine macro to use new types.
6166 (align_labels_log): Redefine macro to use new types.
6167 (align_labels_max_skip): Redefine macro to use new types.
6168 (align_functions_log): Redefine macro to use new types.
6169 (align_loops): Redefine macro to use new types.
6170 (align_jumps): Redefine macro to use new types.
6171 (align_labels): Redefine macro to use new types.
6172 (align_functions): Redefine macro to use new types.
6173 (align_functions_max_skip): Redefine macro to use new types.
6174 (align_loops_value): New macro.
6175 (align_jumps_value): New macro.
6176 (align_labels_value): New macro.
6177 (align_functions_value): New macro.
6178 * function.c (invoke_set_current_function_hook): Propagate
6179 alignment values from flags to global variables default in
6181 * ipa-icf.c (sem_function::equals_wpa): Use
6182 cl_optimization_option_eq instead of memcmp.
6183 * lto-streamer.h (cl_optimization_stream_out): Support streaming
6185 (cl_optimization_stream_in): Likewise.
6186 * optc-save-gen.awk: Support strings in cl_optimization.
6187 * opth-gen.awk: Likewise.
6188 * opts.c (finish_options): Remove error checking of invalid
6190 (MAX_CODE_ALIGN): Remove.
6191 (MAX_CODE_ALIGN_VALUE): Likewise.
6192 (parse_and_check_align_values): New function.
6193 (check_alignment_argument): Likewise.
6194 (common_handle_option): Use check_alignment_argument.
6195 * opts.h (parse_and_check_align_values): Declare.
6196 * toplev.c (init_alignments): Remove.
6197 (read_log_maxskip): New.
6198 (parse_N_M): Likewise.
6199 (parse_alignment_opts): Likewise.
6200 (backend_init_target): Remove usage of init_alignments.
6201 * toplev.h (parse_alignment_opts): Declare.
6202 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
6204 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
6205 * tree.c (cl_option_hasher::equal): New.
6206 * varasm.c: Use new global macros.
6208 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
6210 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
6211 Use a simpler align directive also if MAXSKIP = ALIGN-1.
6212 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6213 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6214 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6215 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6216 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
6217 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
6218 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
6219 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6220 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6221 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6222 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6224 2018-07-04 Martin Liska <mliska@suse.cz>
6225 Jonathan Wakely <jwakely@redhat.com>
6227 * coverage.c: Use correct type.
6228 * doc/invoke.texi: Language correction.
6230 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
6233 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
6234 ENDBRANCH for non-tail call which may return via indirect branch.
6235 * doc/extend.texi: Document indirect_return attribute.
6237 2018-07-03 Martin Sebor <msebor@redhat.com>
6239 PR tree-optimization/86274
6240 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
6242 (format_floating): Correct handling of infinities and NaNs.
6244 2018-07-03 Martin Sebor <msebor@redhat.com>
6246 * print-tree.c (print_real_cst): New function.
6247 (print_node_brief): Call it.
6248 (print_node): Ditto.
6250 2018-07-03 Jeff Law <law@redhat.com>
6252 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
6253 into a single pattern.
6255 * config/h8300/h8300.md (ors code_iterator): New.
6256 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
6257 a single pattern and single splitter.
6258 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
6259 (iorqi3_1, xorqi3_1): Likewise.
6260 (iorqi3, xorqi3 expanders): Similarly.
6262 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
6263 (movmd_internal) into a single pattern using the P mode iterator.
6264 (movmd splitters): Similarly.
6265 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
6266 (movsd splitters): Similarly.
6268 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
6269 ADDB, ADDW and ADDL into a single ADD attribute which selects the
6270 right table based on the size of the operand.
6271 * config/h8300/h8300.md (length_table): Corresponding changes. All
6272 references to "addb", "addw" and "addl" changed to "add".
6273 (btst patterns): Merge two variants into a single pattern.
6274 (tstqi, tsthi): Likewise.
6275 (addhi3_incdec, addsi3_incdec): Likewise.
6276 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
6277 (mulhi3, mulsi3): Likewise.
6278 (udivhi3, udivsi3): Likewise.
6279 (divhi3, divsi3): Likewise.
6280 (andorqi3, andorhi3, andorsi3): Likewise.
6282 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
6285 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
6286 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
6288 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6290 PR tree-optimization/85694
6291 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
6292 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
6294 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
6295 (<u>avg<mode>3_ceil): New patterns.
6297 2018-07-03 David Malcolm <dmalcolm@redhat.com>
6299 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
6300 scan-tree-dump directive.
6301 * gcc.dg/vect/slp-perm-2.c: Likewise.
6302 * gcc.dg/vect/slp-perm-3.c: Likewise.
6303 * gcc.dg/vect/slp-perm-5.c: Likewise.
6304 * gcc.dg/vect/slp-perm-6.c: Likewise.
6305 * gcc.dg/vect/slp-perm-7.c: Likewise.
6306 * gcc.dg/vect/slp-perm-8.c: Likewise.
6308 2018-07-03 Marek Polacek <polacek@redhat.com>
6311 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
6313 2018-07-03 Richard Biener <rguenther@suse.de>
6316 * tree-ssa-structalias.c (find_func_clobbers): Properly
6317 handle indirect calls.
6319 2018-07-03 Jeff Law <law@redhat.com>
6321 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
6322 (shifts): New code iterator.
6323 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
6324 expander. Fix HImode handling on H8/SX.
6325 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
6326 (subqi3, subhi3, subsi3 expanders): Likewise.
6327 (andqi3, andhi3, andsi3 expanders): Likewise.
6328 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
6329 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
6330 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
6331 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
6332 (zero_extendqihi2, zero_extendqisi2): Likewise.
6333 (extendqihi2, extendqisi2): Likewise.
6334 (rotlqi3, rotlhi3, rotlsi3): Likewise.
6335 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
6336 (rotlqi3_1, rotlhi3_1): Likewise.
6337 (logicalhi3_sn, logicalsi3_sn): Likewise.
6338 (logicalhi3, logicalsi3): Likewise.
6340 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6342 * tree-vect-patterns.c (vect_recog_rotate_pattern)
6343 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
6344 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
6345 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
6346 type to append_pattern_def_seq instead of creating a stmt_vec_info
6348 (build_mask_conversion): Likewise. Remove vinfo argument.
6349 (vect_add_conversion_to_patterm): Likewise, renaming to...
6350 (vect_add_conversion_to_pattern): ...this.
6351 (vect_recog_mask_conversion_pattern): Update call to
6352 build_mask_conversion. Pass the vector type to
6353 append_pattern_def_seq here too.
6354 (vect_recog_gather_scatter_pattern): Update call to
6355 vect_add_conversion_to_pattern.
6357 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6359 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
6360 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
6361 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
6362 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
6363 STMT_VINFO_PATTERN_DEF_SEQ to null here.
6364 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
6365 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
6366 append_pattern_def_seq instead of new_pattern_def_seq.
6367 (vect_recog_divmod_pattern): Do both of the above.
6368 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
6371 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6373 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
6374 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
6375 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
6376 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
6377 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
6378 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
6379 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
6380 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
6381 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
6382 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
6383 parameter with a single stmt_vec_info.
6384 (vect_recog_func_ptr): Likewise.
6385 (vect_recog_gather_scatter_pattern): Likewise, folding in...
6386 (vect_try_gather_scatter_pattern): ...this.
6387 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
6388 the stmt_vec_info of the statement to be matched. Don't clear
6389 STMT_VINFO_RELATED_STMT.
6390 (vect_pattern_recog): Update call accordingly.
6392 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6394 PR tree-optimization/85694
6395 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
6396 (uavgM3_ceil): Document new optabs.
6397 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
6398 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
6400 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
6401 (savg_ceil_optab): New optabs.
6402 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
6403 (vect_vect_recog_func_ptrs): Add it.
6404 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
6405 constant directly from the associated lhs.
6407 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6409 * tree-vect-patterns.c (vect_split_statement): New function.
6410 (vect_convert_input): Use it to try to split an existing cast.
6412 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6414 * poly-int.h (print_hex): New function.
6415 * dumpfile.h (dump_dec, dump_hex): Declare.
6416 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
6417 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
6418 min_input_precision, operation_precision and operation_sign.
6419 * tree-vect-patterns.c (vect_get_range_info): New function.
6420 (vect_same_loop_or_bb_p, vect_single_imm_use)
6421 (vect_operation_fits_smaller_type): Delete.
6422 (vect_look_through_possible_promotion): Add an optional
6423 single_use_p parameter.
6424 (vect_recog_over_widening_pattern): Rewrite to use new
6425 stmt_vec_info infomration. Handle one operation at a time.
6426 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
6427 (vect_truncatable_operation_p, vect_set_operation_type)
6428 (vect_set_min_input_precision): New functions.
6429 (vect_determine_min_output_precision_1): Likewise.
6430 (vect_determine_min_output_precision): Likewise.
6431 (vect_determine_precisions_from_range): Likewise.
6432 (vect_determine_precisions_from_users): Likewise.
6433 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
6434 (vect_vect_recog_func_ptrs): Put over_widening first.
6436 (vect_pattern_recog): Call vect_determine_precisions.
6438 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6440 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
6441 statements that have been replaced by further pattern statements.
6442 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
6444 2018-07-03 Richard Biener <rguenther@suse.de>
6446 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
6447 always set *dt. Dump vectype in vectype overload.
6448 * dumpfile.h (dump_gimple_expr): New function.
6449 (dump_gimple_expr_loc): Likewise.
6450 * dumpfile.c (dump_gimple_expr): New function.
6451 (dump_gimple_expr_loc): Likewise.
6453 2018-07-02 Jeff Law <law@redhat.com>
6455 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
6456 the H8/300, H8/300H and H8/S variants into a single pattern.
6457 (movhi_h8300, movqi_h8300hs): Similarly.
6458 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
6459 (QHI mode iterator): New.
6461 * config/h8300/h8300.md: Remove trailing whitespace.
6463 2018-07-02 Jim Wilson <jimw@sifive.com>
6465 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
6466 instead of emit_insn for interrupt returns.
6467 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
6468 (riscv_sret, riscv_uret): Likewise.
6470 2018-07-02 David Malcolm <dmalcolm@redhat.com>
6472 * pretty-print.c (selftest::test_pp_format): Move save and restore
6473 of quotes to class auto_fix_quotes, and add an instance.
6474 * selftest.c: Include "intl.h".
6475 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
6476 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
6477 * selftest.h (selftest::auto_fix_quotes): New class.
6479 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
6481 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
6482 (aarch64_sve_prepare_conditional_op): Remove.
6483 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
6484 Allow aarch64_simd_reg_or_zero as select operand; remove
6485 the aarch64_sve_prepare_conditional_op call.
6486 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
6487 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
6488 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
6489 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
6490 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
6491 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
6492 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
6493 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
6494 and a splitters to match all of the *_any patterns.
6495 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
6497 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
6498 (SVE_COND_FP_BINARY_REV): Remove.
6499 (sve_int_op_rev, sve_fp_op_rev): New.
6500 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
6501 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
6502 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
6503 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
6504 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
6505 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
6506 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
6507 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
6508 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
6510 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
6511 Remove match_dup 1 from the inner unspec.
6512 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
6514 * config/aarch64/aarch64.md (movprfx): New attr.
6515 (length): Default movprfx to 8.
6516 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
6517 (*madd<SVE_I>, *msub<SVE_I): Likewise.
6518 (*<su>mul<SVE_I>3_highpart): Likewise.
6519 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
6520 (*v<ASHIFT><SVE_I>3): Likewise.
6521 (*<su><MAXMIN><SVE_I>3): Likewise.
6522 (*<su><MAXMIN><SVE_F>3): Likewise.
6523 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
6524 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
6525 (*div<SVE_F>4): Likewise.
6527 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
6529 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
6532 2018-07-02 Richard Biener <rguenther@suse.de>
6534 PR tree-optimization/86363
6535 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
6536 memset argument refers to a non-variable address.
6538 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
6540 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
6541 {PLUS,MINUS}_EXPR code to...
6542 (adjust_symbolic_bound): ...here,
6543 (combine_bound): ...here,
6544 (set_value_range_with_overflow): ...and here.
6546 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
6548 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
6550 (extract_range_from_abs_expr): ...here.
6552 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
6554 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
6555 -fno-omit-frame-pointer when not optimizing.
6557 2018-07-02 Martin Liska <mliska@suse.cz>
6560 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
6561 (propagate_nothrow): Likewise.
6563 2018-07-02 Martin Liska <mliska@suse.cz>
6566 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
6568 2018-07-02 David Malcolm <dmalcolm@redhat.com>
6570 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
6571 function in r262149, changing "loc" param from source_location to
6572 const dump_location_t &.
6573 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
6574 declaration, as above.
6576 2018-07-01 Paul Koning <ni1d@arrl.net>
6578 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
6579 -munit-asm, -mgnu-asm, -mdec-asm.
6580 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
6581 (pdp11_output_labelref): New.
6582 (pdp11_output_def): New.
6583 (pdp11_output_addr_vec_elt): New.
6584 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
6585 %# and %@ format codes.
6586 (pdp11_option_override): New.
6587 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
6588 (pdp11_output_ident): New.
6589 (pdp11_asm_named_section): New.
6590 (pdp11_asm_init_sections): New.
6591 (pdp11_file_start): New.
6592 (pdp11_file_end): New.
6593 (output_ascii): Use .ascii/.asciz for -mdec-asm.
6594 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
6595 %o, like %c but octal.
6596 (pdp11_option_override): New.
6597 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
6599 (DATA_SECTION_ASM_OP): Ditto.
6600 (READONLY_DATA_SECTION_ASM_OP): New.
6601 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
6602 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
6603 (ASM_OUTPUT_LABELREF): Ditto.
6604 (ASM_OUTPUT_DEF): Ditto.
6605 (ASM_OUTPUT_EXTERNAL): New.
6606 (ASM_OUTPUT_SOURCE_FILENAME): New.
6607 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
6608 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
6609 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
6610 %# and %@ format codes.
6611 * config/pdp11/pdp11.opt (mgnu-asm): New.
6612 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
6613 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
6614 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
6616 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
6618 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
6619 dereferencing path[] beyond its length.
6621 (debug_all_paths): New.
6622 (rewire_first_differing_edge): New.
6623 (adjust_paths_after_duplication): New.
6624 (duplicate_thread_path): Call adjust_paths_after_duplication.
6626 (thread_through_all_blocks): Add new argument to
6627 duplicate_thread_path.
6629 2018-06-30 Jim Wilson <jimw@sifive.com>
6631 * config/riscv/predicates.md (p2m1_shift_operand): New.
6632 (high_mask_shift_operand): New.
6633 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
6634 pattern using p2m1_shift_operand.
6635 (lshsi3_zero_extend_3+2): New combiner pattern using
6636 high_mask_shift_operand.
6638 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6640 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
6642 (vect_recog_rotate_pattern): ...here.
6643 (vect_convert_input): Try to insert casts of invariants in the
6645 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
6646 preheader to be empty.
6648 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6650 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
6651 vector type. If given, install it in the new statement's
6653 (vect_element_precision): New function.
6654 (vect_unpromoted_value): New struct.
6655 (vect_unpromoted_value::vect_unpromoted_value): New function.
6656 (vect_unpromoted_value::set_op): Likewise.
6657 (vect_look_through_possible_promotion): Likewise.
6658 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
6659 (vect_widened_op_tree, vect_convert_input): Likewise.
6660 (vect_convert_inputs, vect_convert_output): Likewise.
6661 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
6662 to handle the optional cast of the multiplication result and
6663 vect_widened_op_tree to detect the widened multiplication itself.
6664 Do not require the input and output of promotion casts to have
6665 the same sign, but base the signedness of the operation on the
6666 input rather than the result. If the pattern includes two
6667 promotions, check that those promotions have the same sign.
6668 Do not restrict the MULT_EXPR handling to a double-width result;
6669 handle quadruple-width results and wider. Use vect_convert_inputs
6670 to convert the inputs to the common type.
6671 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
6672 to handle the optional cast of the ABS result. Also allow a sign
6673 change or a sign extension between the ABS and MINUS.
6674 Use vect_widened_op_tree to detect the widened subtraction and use
6675 vect_convert_inputs to convert the inputs to the common type.
6676 (vect_handle_widen_op_by_const): Delete.
6677 (vect_recog_widen_op_pattern): New function.
6678 (vect_recog_widen_mult_pattern): Use it.
6679 (vect_recog_widen_shift_pattern): Likewise.
6680 (vect_recog_widen_sum_pattern): Use
6681 vect_look_through_possible_promotion to handle the promoted
6684 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6686 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
6687 the containing gimple_seq *.
6688 * gimple-iterator.h (gsi_for_stmt): Declare it.
6689 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
6690 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
6691 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
6692 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
6693 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
6695 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
6697 (vect_mark_pattern_stmts): ...here. Handle cases in which the
6698 statement being replaced is part of an existing pattern
6699 definition sequence, inserting the new pattern statements before
6701 (vect_pattern_recog_1): Don't return a bool. If the statement
6702 is already part of a pattern, instead apply pattern matching
6703 to the pattern definition statements. Don't clear the
6704 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
6705 (vect_pattern_recog): Don't break after the first match;
6706 continue processing the pattern definition statements instead.
6707 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
6709 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6711 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6712 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
6713 (vect_recog_widen_sum_pattern): Use it.
6715 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6717 * tree-vect-loop.c (vectorizable_reduction): Assert that the
6718 phi is not a pattern statement and has not been replaced by
6719 a pattern statement.
6720 * tree-vect-patterns.c (type_conversion_p): Don't check
6721 STMT_VINFO_IN_PATTERN_P.
6722 (vect_recog_vector_vector_shift_pattern): Likewise.
6723 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
6724 the pattern statement rather than the original statement; check
6725 directly for a WIDEN_MULT_EXPR here.
6726 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
6727 vect_is_simple_use to return the pattern statement rather
6728 than the original statement; use is_pattern_stmt_p to check
6729 for such a pattern statement.
6730 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
6731 to return the pattern statement rather than the original statement;
6732 don't do the same transformation here.
6733 (vect_is_simple_use): If the defining statement has been replaced
6734 by a pattern statement, return the pattern statement instead.
6735 Remove the corresponding (local) transformation from the vectype
6738 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6740 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
6741 end and default to null.
6742 * tree-vect-loop.c (vect_create_epilog_for_reduction)
6743 (vectorizable_reduction): Update calls accordingly, dropping the
6744 gimple ** argument if the passed-back statement isn't needed.
6745 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
6746 (vect_recog_rotate_pattern): Likewise.
6747 (vect_recog_mask_conversion_pattern): Likewise.
6748 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6749 (vect_mask_constant_operand_p): Likewise.
6750 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
6751 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
6752 (get_group_load_store_type, get_load_store_type): Likewise.
6753 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
6754 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
6755 (vectorizable_conversion, vectorizable_assignment): Likewise.
6756 (vectorizable_shift, vectorizable_operation): Likewise.
6757 (vectorizable_store, vect_is_simple_cond): Likewise.
6758 (vectorizable_condition, vectorizable_comparison): Likewise.
6759 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
6760 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
6761 and move it to the end. Cope with null def_stmt_outs.
6763 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
6765 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
6767 2018-06-29 Jeff Law <law@redhat.com>
6769 * config/v850/v850.c (v850_legitimate_address_p): Handle large
6770 displacements for TARGET_V850E2V3 and newer.
6771 (TARGET_LRA_P): Remove. Defaults to LRA now.
6772 * config/v850/v850.md (sign23byte_load): Remove.
6773 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
6774 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
6776 2018-06-29 Martin Liska <mliska@suse.cz>
6779 * coverage.c (coverage_init): Mangle full path name.
6780 * doc/invoke.texi: Document the change.
6781 * gcov-io.c (mangle_path): New.
6782 * gcov-io.h (mangle_path): Likewise.
6783 * gcov.c (mangle_name): Use mangle_path for path mangling.
6785 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6787 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
6788 if starting source register is not even.
6790 2018-06-29 Martin Liska <mliska@suse.cz>
6792 PR tree-optimization/86263
6793 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
6794 Make edge redirection.
6796 2018-06-29 David Malcolm <dmalcolm@redhat.com>
6798 * dumpfile.c (dump_loc): Add indentation based on scope depth.
6799 (dump_scope_depth): New variable.
6800 (get_dump_scope_depth): New function.
6801 (dump_begin_scope): New function.
6802 (dump_end_scope): New function.
6803 * dumpfile.h (get_dump_scope_depth): New declaration.
6804 (dump_begin_scope): New declaration.
6805 (dump_end_scope): New declaration.
6806 (class auto_dump_scope): New class.
6807 (AUTO_DUMP_SCOPE): New macro.
6808 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
6811 2018-06-29 Richard Biener <rguenther@suse.de>
6813 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
6814 compute_all_dependences succeeds.
6815 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
6816 exceed --param loop-max-datarefs-for-datadeps.
6818 2018-06-29 Jakub Jelinek <jakub@redhat.com>
6820 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
6822 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
6825 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
6826 alternative 0 in preferred_for_speed attribute.
6828 2018-06-28 Paul Koning <ni1d@arrl.net>
6830 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
6831 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
6832 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
6833 * config/pdp11/pdp11.md: Correct "length" attribute calculation
6834 for shift insn patterns.
6836 2018-06-28 David Malcolm <dmalcolm@redhat.com>
6838 * cgraph.c (cgraph_node::get_body): Replace assignments to
6839 "dump_file" with calls to set_dump_file.
6840 * dumpfile.c (alt_dump_file): Make static, and group with...
6841 (alt_flags): ...this definition.
6842 (dumps_are_enabled): New variable.
6843 (refresh_dumps_are_enabled): New function.
6844 (set_dump_file): New function.
6845 (set_alt_dump_file): New function.
6846 (gcc::dump_manager::dump_start): Replace assignments to
6847 "dump_file" and "alt_dump_file" with calls to set_dump_file and
6849 (gcc::dump_manager::dump_finish): Likewise.
6850 * dumpfile.h (alt_dump_file): Delete decl.
6851 (dumps_are_enabled): New variable decl.
6852 (set_dump_file): New function decl.
6853 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
6855 * tree-nested.c (lower_nested_functions): Replace assignments to
6856 "dump_file" with calls to set_dump_file.
6858 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
6860 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
6861 goto_locus of each outgoing edge of each basic block.
6863 2018-06-28 Richard Biener <rguenther@suse.de>
6865 * dwarf2out.c (decl_scope_table): Remove.
6866 (push_decl_scope): Likewise.
6867 (pop_decl_scope): Likewise.
6868 (gen_type_die_for_member): Do not call push/pop_decl_scope.
6869 (gen_struct_or_union_type_die): Likewise.
6870 (gen_tagged_type_die): Likewise.
6871 (dwarf2out_init): Do not initialize decl_scope_table.
6872 (dwarf2out_c_finalize): Do not free it.
6874 2018-06-28 Richard Biener <rguenther@suse.de>
6876 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
6877 deciding whether to not re-use a DIE.
6879 2018-06-28 Richard Biener <rguenther@suse.de>
6881 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
6882 DW_AT_abstract_origin attribute.
6884 2018-06-28 Martin Liska <mliska@suse.cz>
6886 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
6887 Use newly introduced constants.
6888 * tree-switch-conversion.h (struct jump_table_cluster):
6889 Define max_ratio_for_size and max_ratio_for_speed.
6891 2018-06-28 Martin Liska <mliska@suse.cz>
6893 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
6894 Add new checking assert to catch invalid state.
6895 (jump_table_cluster::can_be_handled): Handle single case
6897 (jump_table_cluster::is_beneficial): Bail out for such case.
6898 (bit_test_cluster::find_bit_tests):
6899 Add new checking assert to catch invalid state.
6900 (bit_test_cluster::can_be_handled): Handle single case
6902 (bit_test_cluster::is_beneficial): Bail out for such case.
6903 (switch_decision_tree::analyze_switch_statement):
6906 2018-06-28 Martin Liska <mliska@suse.cz>
6908 * common.opt: Introduce -completion option.
6909 * gcc.c (driver_handle_option): Handle it.
6910 (driver::main): Print completions if completion
6912 * opt-suggestions.c (option_proposer::get_completions):
6914 (option_proposer::suggest_completion): Likewise.
6915 (option_proposer::find_param_completions): Likewise.
6916 (verify_autocompletions): Likewise.
6917 (test_completion_valid_options): Likewise.
6918 (test_completion_valid_params): Likewise.
6919 (in_completion_p): Likewise.
6920 (empty_completion_p): Likewise.
6921 (test_completion_partial_match): Likewise.
6922 (test_completion_garbage): Likewise.
6923 (opt_proposer_c_tests): Likewise.
6924 * opt-suggestions.h: Declare new functions.
6925 * opts.c (common_handle_option): Handle OPT__completion_.
6926 * selftest-run-tests.c (selftest::run_tests): Add
6927 opt_proposer_c_tests.
6928 * selftest.c (assert_str_startswith): New.
6929 * selftest.h (assert_str_startswith): Likewise.
6930 (opt_proposer_c_tests): New.
6931 (ASSERT_STR_STARTSWITH): Likewise.
6933 2018-06-28 Martin Liska <mliska@suse.cz>
6935 * Makefile.in: Add opt-suggestions.o.
6936 * gcc-main.c: Include opt-suggestions.h.
6937 * gcc.c (driver::driver): Likewise.
6938 (driver::~driver): Remove m_option_suggestions.
6939 (driver::build_option_suggestions): Moved to option_proposer.
6940 (driver::suggest_option): Likewise.
6941 (driver::handle_unrecognized_options): Use option_proposer.
6942 * gcc.h (class driver): Add new memver m_option_proposer.
6943 * opt-suggestions.c: New file.
6944 * opt-suggestions.h: New file.
6946 2018-06-28 Martin Liska <mliska@suse.cz>
6948 * vec.h (class auto_string_vec): New (moved from auto_argvec).
6949 (auto_string_vec::~auto_string_vec): Likewise.
6951 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
6953 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
6954 prevent_decl_creation_for_types fields up and add reset_location field.
6955 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
6956 statement if id->reset_location is true.
6957 (copy_edges_for_bb): Do not set goto_locus on the new edges if
6958 id->reset_location is true.
6959 (copy_phis_for_bb): Force input_location on the arguments if
6960 id->reset_location is true.
6961 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
6962 is set on the function to be inlined.
6964 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
6966 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
6968 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
6970 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
6971 registers for Pmode.
6972 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
6973 hard registers for the clobbered pseudo.
6975 2018-06-27 Paul Koning <ni1d@arrl.net>
6977 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
6978 mutually exclusive options.
6979 * config/pdp11/constraints.md (h): New constraint.
6980 (O): Update definition to match shift code generation.
6981 (D): New constraint.
6982 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
6984 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
6986 (output_jump): Change arguments.
6987 (pdp11_fixed_cc_regs): New function.
6988 (pdp11_cc_mode): Ditto.
6989 (pdp11_expand_shift): Ditto.
6990 (pdp11_assemble_shift): Ditto.
6991 (pdp11_small_shift): Ditto.
6992 (pdp11_branch_cost): Remove.
6993 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
6995 (pdp11_register_move_cost): Update for CC registers.
6996 (pdp11_rtx_costs): Add case for LSHIFTRT.
6997 (pdp11_output_jump): Add CCNZ mode conditional branches.
6998 (notice_update_cc_on_set): Remove.
6999 (pdp11_cc_mode): New function.
7000 (simple_memory_operand): Correct pre/post decrement case.
7001 (no_side_effect_operand): New function.
7002 (pdp11_regno_reg_class): Add CC_REGS class.
7003 (pdp11_fixed_cc_regs): New function.
7004 (pdp11_small_shift): New function.
7005 (pdp11_expand_shift): New function to expand shift insns.
7006 (pdp11_assemble_shift): New function to output shifts.
7007 (pdp11_branch_cost): Remove.
7008 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
7009 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
7010 (WCHAR_TYPE): Ditto.
7011 (PTRDIFF_TYPE): Ditto.
7012 (ADJUST_INSN_LENGTH): New macro.
7013 (FIXED_REGISTERS): Add CC registers.
7014 (CALL_USED_REGISTERS): Ditto.
7016 (REG_CLASS_NAMES): Ditto.
7017 (REG_CLASS_CONTENTS): Ditto.
7018 (SELECT_CC_MODE): Use new function.
7019 (TARGET_FLAGS_REGNUM): New macro.
7020 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
7021 (cc0_reg_rtx): Remove.
7022 (CC_STATUS_MDEP): Remove.
7023 (CC_STATUS_MDEFP_INIT): Remove.
7024 (CC_IN_FPU): Remove.
7025 (NOTICE_UPDATE_CC): Remove.
7026 (REGISTER_NAMES): Add CC registers.
7027 (BRANCH_COST): Change to constant 1.
7028 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
7030 * config/pdp11/pdp11.opt (mbcopy): Remove.
7031 (mbcopy-builtin): Remove.
7032 (mbranch-cheap): Remove.
7033 (mbranch-expensive): Remove.
7034 * config/pdp11/predicates.md (expand_shift_operand): Update to
7035 match shift code generation.
7036 (ccnz_operator): New predicate.
7037 * doc/invoke.texi (PDP-11 Options): Remove deleted options
7038 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
7039 Remove non-existent option -mabshi, -mno-abshi. Document mutually
7041 * doc/md.texi (PDP-11): Document new D and h constraints. Update
7042 description of O constraint.
7044 2018-06-27 Jeff Law <law@redhat.com>
7045 Austin Law <austinklaw@gmail.com>
7047 * config/v850/v850.md (addsi3_set_flags): New pattern.
7048 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
7049 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
7050 (zero_extendhisi2_v850_set_flags): Likewise.
7051 (zero_extendqisi2_v850_set_flags): Likewise.
7052 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
7053 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
7054 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
7056 * config/v850/v850-protos.h (notice_update_cc): Remove.
7057 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
7058 (v850_print_operand): Handle 'D' and "d".
7059 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
7060 Add handling of arithmetic/logical operations compared against zero.
7061 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
7062 Do not look at v850_compare_op, instead get mode from last argument.
7063 (v850_gen_compare): Remove
7064 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
7065 after reload for prologue insns.
7066 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
7068 (construct_save_jarl): Likewise.
7069 (TARGET_FLAGS_REGNUM): Define.
7070 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
7071 (NOTICE_UPDATE_CC): Remove.
7072 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
7073 than cc0. Conditionalize on reload_completed.
7074 (cmpsi_insn, setfcc_insn): Likewise.
7075 (tst1 splitter): Turn into define_and_split which sets the flags
7077 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
7078 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
7079 (cstoresf4, cstoredf4): Clobber the flags.
7080 (cmpsi, cmpsf, cmpdf): Remove expanders.
7081 (setf_insn): Remove pattern.
7082 (addsi3): Turn into define_and_split which clobbers the flags after
7083 reload and a suitable pattern (addsi3_clobber_flags) for use after
7085 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
7086 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
7087 (ashrsi3, ashrsi3_v850e2): Likewise.
7088 (bins): Clobber the flags.
7089 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
7090 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
7091 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
7092 (call_value_internal_short, call_value_internal_long): Likewise.
7093 (callt_save_interrupt, callt_return_interrupt): Likewise.
7094 (save_interrupt, return_interrupt): Likewise.
7095 (callt_save_all_interrupt, save_all_interrupt): Likewise.
7096 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
7097 (restore_all_interrupt, _restore_all_interrupt): Likewise.
7098 (All FP comparisons): Only allow after reload has completed.
7100 (divh, divhu): Tweak output template.
7101 (branch_z_normal, branch_z_invert): Remove
7102 (branch_nz_normal, branch_nz_invert): Likewise.
7103 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
7105 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
7106 * config/v850/v850.c (notice_update_cc): Remove.
7107 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
7108 (CC_NO_CARRY): Likewise.
7109 (NOTICE_UPDATE_CC): Define to nothing.
7110 * config/v850/v850.md: Remove block comment on cc0 handling
7111 Remove "cc" attribute from all patterns. Remove cc_status handling
7112 from all patterns. Minor formatting fixes.
7114 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7116 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
7117 (cortex-a76.cortex-a55): Likewise.
7118 * config/aarch64/aarch64-tune.md: Regenerate.
7119 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
7120 cortex-a76.cortex-a55.
7122 2018-06-27 Jeff Law <law@redhat.com>
7124 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
7125 (MULTILIB_DIRNAMES): Similarly.
7127 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
7129 * gimple.h (gimple_return_retbnd): Delete.
7130 (gimple_return_set_retbnd): Likewise.
7131 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
7132 gimple_return_set_retbnd.
7133 * gimple-pretty-print.c (dump_gimple_return): Remove call to
7134 gimple_return_retbnd and adjust.
7135 * tree-inline.h (struct copy_body_data): Remove retbnd field.
7136 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
7137 Explicitly return NULL in a couple more cases. Move assertion
7138 on debug statements and remove unreachable code.
7139 (reset_debug_binding): Do not test id->retbnd.
7140 (expand_call_inline): Do not set it.
7142 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
7144 * configure.ac: Add --disable-gcov option.
7145 * configure: Regenerate.
7146 * Makefile.in: Honour @enable_gcov@.
7147 * doc/install.texi: Document --disable-gcov.
7149 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7151 * config/arm/arm-cpus.in (cortex-a76): New entry.
7152 (cortex-a76.cortex-a55): Likewise.
7153 * config/arm/arm-tables.opt: Regenerate.
7154 * config/arm/arm-tune.md: Likewise.
7155 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
7156 * doc/invoke.texi (ARM Options): Document cortex-a76 and
7157 cortex-a76.cortex-a55.
7159 2018-06-27 Tamar Christina <tamar.christina@arm.com>
7162 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
7164 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
7166 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
7168 (EPILOGUE_USES): Likewise.
7170 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
7172 * tree-inline.c (remap_location): New function extracted from...
7173 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
7174 (copy_phis_for_bb): ...here. Call remap_location.
7175 (copy_cfg_body): Adjust call to copy_edges_for_bb.
7177 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
7179 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
7180 unaligned vsx for 16B memset.
7182 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
7185 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
7186 ieee128_float_type_node to long_double_type_node unless
7187 TARGET_LONG_DOUBLE_128 is set.
7189 2018-06-26 David Malcolm <dmalcolm@redhat.com>
7191 * cfgloop.c (get_loop_location): Convert return type from
7192 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
7193 by implicit construction from rtx_insn *, and using
7194 dump_user_location_t::from_function_decl for the fallback case.
7195 * cfgloop.h (get_loop_location): Convert return type from
7196 location_t to dump_user_location_t.
7197 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
7198 dump_printf_loc to pass in a dump_location_t rather than a
7199 location_t, via the gimple stmt.
7200 * coverage.c (get_coverage_counts): Update calls to
7201 dump_printf_loc to pass in dump_location_t rather than a
7203 * doc/optinfo.texi (Dump types): Convert example of
7204 dump_printf_loc from taking "locus" to taking "insn". Update
7205 description of the "_loc" calls to cover dump_location_t.
7206 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
7208 (dump_user_location_t::dump_user_location_t): New constructors,
7209 from gimple *stmt and rtx_insn *.
7210 (dump_user_location_t::from_function_decl): New function.
7211 (dump_loc): Make static.
7212 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
7213 const dump_location_t &.
7214 (dump_generic_expr_loc): Delete.
7215 (dump_printf_loc): Convert param "loc" from location_t to
7216 const dump_location_t &.
7217 (selftest::test_impl_location): New function.
7218 (selftest::dumpfile_c_tests): New function.
7219 * dumpfile.h: Include "profile-count.h".
7220 (class dump_user_location_t): New class.
7221 (struct dump_impl_location_t): New struct.
7222 (class dump_location_t): New class.
7223 (dump_printf_loc): Convert 2nd param from source_location to
7224 const dump_location_t &.
7225 (dump_generic_expr_loc): Delete.
7226 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
7227 const dump_location_t &.
7228 * gimple-fold.c (fold_gimple_assign): Update call to
7229 dump_printf_loc to pass in a dump_location_t rather than a
7230 location_t, via the gimple stmt.
7231 (gimple_fold_call): Likewise.
7232 * gimple-loop-interchange.cc
7233 (loop_cand::analyze_iloop_reduction_var): Update for change to
7234 check_reduction_path.
7235 (tree_loop_interchange::interchange): Update for change to
7237 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
7238 change in return-type of find_loop_location.
7239 (graphite_regenerate_ast_isl): Likewise.
7240 * graphite-optimize-isl.c (optimize_isl): Likewise.
7241 * graphite.c (graphite_transform_loops): Likewise.
7242 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
7243 pass in a dump_location_t rather than a location_t, via the
7245 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7246 * ipa.c (walk_polymorphic_call_targets): Likewise.
7247 * loop-unroll.c (report_unroll): Convert "locus" param from
7248 location_t to dump_location_t.
7249 (decide_unrolling): Update for change to get_loop_location's
7251 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
7252 location_t to dump_user_location_t.
7253 (grid_find_single_omp_among_assignments_1): Updates calls to
7254 dump_printf_loc to pass in a dump_location_t rather than a
7255 location_t, via the gimple stmt.
7256 (grid_parallel_clauses_gridifiable): Convert "tloc" from
7257 location_t to dump_location_t. Updates calls to dump_printf_loc
7258 to pass in a dump_location_t rather than a location_t, via the
7260 (grid_inner_loop_gridifiable_p): Likewise.
7261 (grid_dist_follows_simple_pattern): Likewise.
7262 (grid_gfor_follows_tiling_pattern): Likewise.
7263 (grid_target_follows_gridifiable_pattern): Likewise.
7264 (grid_attempt_target_gridification): Convert initialization
7265 of local "grid" from memset to zero-initialization; FIXME: does
7266 this require C++11? Update call to dump_printf_loc to pass in a
7267 optinfo_location rather than a location_t, via the gimple stmt.
7268 * profile.c (read_profile_edge_counts): Updates call to
7269 dump_printf_loc to pass in a dump_location_t rather than a
7271 (compute_branch_probabilities): Likewise.
7272 * selftest-run-tests.c (selftest::run_tests): Call
7274 * selftest.h (dumpfile_c_tests): New decl.
7275 * tree-loop-distribution.c (pass_loop_distribution::execute):
7276 Update for change in return type of find_loop_location.
7277 * tree-parloops.c (parallelize_loops): Likewise.
7278 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
7279 "locus" from location_t to dump_user_location_t.
7280 (canonicalize_loop_induction_variables): Likewise.
7281 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
7282 for change in return type of find_loop_location.
7283 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
7284 to dump_printf_loc to pass in a dump_location_t rather than a
7285 location_t, via the stmt.
7286 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
7288 * tree-vect-loop-manip.c (find_loop_location): Convert return
7289 type from source_location to dump_user_location_t.
7290 (vect_do_peeling): Update for above change.
7291 (vect_loop_versioning): Update for change in type of
7293 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
7294 from location_t to dump_user_location_t.
7295 (vect_estimate_min_profitable_iters): Update for change in type
7297 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
7298 location_t to dump_location_t.
7299 (vect_slp_bb): Update for change in type of vect_location.
7300 * tree-vectorizer.c (vect_location): Convert from source_location
7301 to dump_user_location_t.
7302 (try_vectorize_loop_1): Update for change in vect_location's type.
7303 (vectorize_loops): Likewise.
7304 (increase_alignment): Likewise.
7305 * tree-vectorizer.h (vect_location): Convert from source_location
7306 to dump_user_location_t.
7307 (find_loop_location): Convert return type from source_location to
7308 dump_user_location_t.
7309 (check_reduction_path): Convert 1st param from location_t to
7310 dump_user_location_t.
7311 * value-prof.c (check_counter): Update call to dump_printf_loc to
7312 pass in a dump_user_location_t rather than a location_t; update
7313 call to error_at for change in type of "locus".
7314 (check_ic_target): Update call to dump_printf_loc to
7315 pass in a dump_user_location_t rather than a location_t, via the
7318 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
7320 * config/s390/s390.h (enum processor_flags): Do not use
7321 default tune parameter when -march was specified.
7323 2018-06-26 Jakub Jelinek <jakub@redhat.com>
7326 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
7327 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
7330 2018-06-26 Richard Biener <rguenther@suse.de>
7332 PR tree-optimization/86287
7334 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
7335 (vect_analyze_loop): Initialize n_stmts.
7337 2018-06-26 Richard Biener <rguenther@suse.de>
7340 * fold-const.c (fold_convertible_p): Pointer extension
7343 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
7346 * dwarf2out.c (loc_list_has_views): Adjust comments.
7347 (dw_loc_list): Split single cross-partition range with
7350 2018-06-25 Jeff Law <law@redhat.com>
7352 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
7353 on -mbig-switch by default.
7355 * config/v850/predicates.md (const_float_1_operand): Fix match_code
7357 (const_float_0_operand): Remove unused predicate.
7358 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
7359 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
7360 (recipsf2): New expander. Original pattern now called
7362 (recipdf2, recipdf2_insn): Similarly.
7363 (rsqrtsf2, rsqrtsf2_insn): Similarly
7364 (rsqrtdf2, rsqrtdf2_insn): Similarly
7366 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
7368 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
7369 Simplify logic for FreeBSD (twice).
7371 2018-06-25 Martin Sebor <msebor@redhat.com>
7373 PR tree-optimization/86204
7374 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
7375 a strnlen result if it's less than the length of the string.
7377 2018-06-25 Martin Sebor <msebor@redhat.com>
7379 PR tree-optimization/85700
7380 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
7381 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
7382 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
7384 2018-06-25 Martin Sebor <msebor@redhat.com>
7386 * doc/extend.texi (Zero-length arrays): Update and clarify.
7388 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
7390 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
7391 added IEEE/IBM long double multilib support on PowerPC little
7392 endian Linux systems.
7393 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
7394 (MULTILIB_DEFAULTS): Likewise.
7395 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7397 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
7398 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
7399 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
7401 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
7404 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
7405 (REORDER_45): Likewise.
7407 2018-06-25 Jeff Law <law@redhat.com>
7409 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
7410 dividend to 32 bits. Adjust length.
7411 (udivmodhi4): Cleanup output template. Fix length.
7413 2018-06-25 Carl Love <cel@us.ibm.com>
7415 * config/rs6000/vsx.md: Change word selector to prefered location.
7417 2018-06-25 Richard Biener <rguenther@suse.de>
7419 PR tree-optimization/86304
7420 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
7421 epilogue-if-converted loops as well.
7423 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
7425 * lto-section-out.c (lto_begin_section): Do not print section
7426 name for noaddr and unnumbered dumps.
7428 2018-06-25 Richard Biener <rguenther@suse.de>
7430 * tree-vectorizer.h (struct vec_info_shared): New structure
7431 with parts split out from struct vec_info and loop_nest from
7432 struct _loop_vec_info.
7433 (struct vec_info): Adjust accordingly.
7434 (struct _loop_vec_info): Likewise.
7435 (LOOP_VINFO_LOOP_NEST): Adjust.
7436 (LOOP_VINFO_DATAREFS): Likewise.
7437 (LOOP_VINFO_DDRS): Likewise.
7438 (struct _bb_vec_info): Likewise.
7439 (BB_VINFO_DATAREFS): Likewise.
7440 (BB_VINFO_DDRS): Likewise.
7441 (struct _stmt_vec_info): Add dr_aux member.
7442 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
7443 (DR_MISALIGNMENT_UNINITIALIZED): New.
7444 (set_dr_misalignment): Adjust.
7445 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
7446 (vect_analyze_loop): Adjust prototype.
7447 (vect_analyze_loop_form): Likewise.
7448 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
7449 Compute dependences lazily.
7450 (vect_record_base_alignments): Use shared datarefs/ddrs.
7451 (vect_verify_datarefs_alignment): Likewise.
7452 (vect_analyze_data_refs_alignment): Likewise.
7453 (vect_analyze_data_ref_accesses): Likewise.
7454 (vect_analyze_data_refs): Likewise.
7455 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
7456 constructor parameter for shared part.
7457 (vect_analyze_loop_form): Pass in shared part and adjust.
7458 (vect_analyze_loop_2): Pass in storage for the number of
7459 stmts. Move loop nest finding to the caller. Compute
7461 (vect_analyze_loop): Pass in shared part.
7462 (vect_transform_loop): Verify shared datarefs are unchanged.
7463 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
7464 constructor parameter for shared part.
7465 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
7466 (vect_slp_bb): Verify shared datarefs are unchanged before
7468 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
7470 (new_stmt_vec_info): Initialize DR_AUX misalignment to
7471 DR_MISALIGNMENT_UNINITIALIZED.
7472 * tree-vectorizer.c (vec_info::vec_info): Add constructor
7473 parameter for shared part.
7474 (vec_info::~vec_info): Adjust.
7475 (vec_info_shared::vec_info_shared): New.
7476 (vec_info_shared::~vec_info_shared): Likewise.
7477 (vec_info_shared::save_datarefs): Likewise.
7478 (vec_info_shared::check_datarefs): Likewise.
7479 (try_vectorize_loop_1): Construct shared part live for analyses
7480 of a single loop for multiple vector sizes.
7481 * tree-parloops.c (gather_scalar_reductions): Adjust.
7483 2018-06-25 Richard Biener <rguenther@suse.de>
7485 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
7486 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
7487 (vect_analyze_data_refs): Remove similar code from here and
7488 simplify accordingly.
7490 2018-06-25 Richard Biener <rguenther@suse.de>
7492 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
7493 for reverse storage order accesses rather than asserting
7494 they cannot happen here.
7496 2018-06-25 Tom de Vries <tdevries@suse.de>
7499 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
7500 Use data16 instead of .byte for insn prefix.
7502 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
7505 * parser.c (make_char_string_pack): Pass this literal chars
7506 through cpp_interpret_string.
7507 (cp_parser_userdef_numeric_literal): Check the result of
7508 make_char_string_pack.
7510 2018-06-24 Maya Rashish <coypu@sdf.org>
7512 * ginclude/stddef.h: Simplify conditions around avoiding
7513 re-definition of __size_t.
7515 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
7517 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
7518 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
7520 2018-06-22 Maya Rashish <coypu@sdf.org>
7522 * doc/invoke.texi (mno-fancy-math-387): Update for changes
7523 made to OpenBSD and NetBSD through the years.
7525 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7527 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
7528 behavior of vec_pack (vector double, vector double) to match
7529 behavior of vec_float2 (vector double, vector double).
7531 2018-06-22 Olivier Hainque <hainque@adacore.com>
7533 * gimplify.c (gimplify_function_tree): Prevent creation
7534 of a trampoline for the address of the current function
7535 passed to entry/exit instrumentation hooks.
7537 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
7540 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
7543 2018-06-22 Martin Liska <mliska@suse.cz>
7545 PR tree-optimization/86263
7546 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7547 Bail out if is_enabled is false.
7548 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
7550 (jump_table_cluster::is_enabled): New function.
7552 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
7554 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
7555 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
7556 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
7557 (lto_input_ts_binfo_tree_pointers): Likewise.
7558 * tree-streamer-out.c (streamer_write_tree_bitfields,
7559 write_ts_binfo_tree_pointers): Likewise.
7560 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
7562 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
7564 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
7566 2018-06-22 Martin Liska <mliska@suse.cz>
7568 * symbol-summary.h (get): Make it pure and inline move
7569 functionality from ::get function.
7570 (get): Remove and inline into ::get and ::get_create.
7571 (get_create): Move code from ::get function.
7573 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7576 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
7577 -x assembler-with-cpp.
7579 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7581 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
7582 _FILE_OFFSET_BITS=64 for C++.
7584 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
7586 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
7587 conversion insn that shows up when pr85657-3.c is compiled using
7588 IEEE 128-bit long double.
7589 (neg<mode>2_internal): Use the correct mode to check whether the
7590 mode is IBM extended.
7591 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
7592 multiply and divide external functions from being created more
7595 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
7597 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
7599 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
7600 the edge can be forwarded.
7601 (cfg_layout_merge_blocks): Likewise.
7603 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
7605 * except.c (finish_eh_generation): Commit edge insertions only after
7606 the EH edges have been redirected from post-landing to landing pads.
7608 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
7610 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
7611 create_tmp_var_for to create the FRAME decl.
7612 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
7614 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
7616 * tree-inline.c (copy_edges_for_bb): Minor tweak.
7617 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
7618 debug statement when resetting its value.
7619 (expand_call_inline): Copy the locus of the call onto the assignment
7620 of the return value, if any. Use local variable in more cases.
7622 2018-06-21 Martin Liska <mliska@suse.cz>
7624 * ipa-pure-const.c (propagate_nothrow): Use
7625 funct_state_summaries->get.
7626 (dump_malloc_lattice): Likewise.
7627 (propagate_malloc): Likewise.
7629 2018-06-21 Richard Biener <rguenther@suse.de>
7631 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
7632 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
7633 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
7635 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
7636 BLOCK_ABSTRACT_ORIGIN unconditionally.
7638 2018-06-21 David Malcolm <dmalcolm@redhat.com>
7640 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
7642 * ipa-reference.c (ipa_reference_c_finalize): Delete
7643 ipa_ref_opt_sum_summaries and set it to NULL.
7645 2018-06-21 Tom de Vries <tdevries@suse.de>
7647 PR tree-optimization/85859
7648 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
7649 test with comment from bb_no_side_effects_p.
7651 2018-06-21 Richard Biener <rguenther@suse.de>
7653 PR tree-optimization/86232
7654 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
7655 max for constant niter.
7657 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
7659 * config/aarch64/aarch64-simd.md
7660 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
7662 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
7664 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
7665 Make opernads of the unspec commutative.
7667 2018-06-21 Richard Biener <rguenther@suse.de>
7669 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
7670 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
7671 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
7672 (vect_analyze_data_ref_dependence): Re-order checks to deal with
7674 (vect_record_base_alignments): Do not record base alignment
7675 for gathers or scatters.
7676 (vect_compute_data_ref_alignment): Drop return value that is always
7677 true. Bail out early for gathers or scatters.
7678 (vect_enhance_data_refs_alignment): Bail out early for gathers
7680 (vect_find_same_alignment_drs): Likewise.
7681 (vect_analyze_data_refs_alignment): Remove dead code.
7682 (vect_slp_analyze_and_verify_node_alignment): Likewise.
7683 (vect_analyze_data_refs): For possible gathers or scatters do
7684 not create an alternate DR, just check their possible validity
7685 and mark them. Adjust DECL_NONALIASED handling to not rely
7687 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
7688 update inits of gathers or scatters.
7689 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
7690 Also copy gather/scatter flag to pattern vinfo.
7692 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
7694 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
7695 behavior of vec_packsu (vector unsigned long long, vector unsigned
7696 long long) to match behavior of vec_packs with same signature.
7698 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
7699 Thomas Schwinge <thomas@codesourcery.com>
7700 Cesar Philippidis <cesar@codesourcery.com>
7702 * gimplify.c (gimplify_scan_omp_clauses): Add support for
7703 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
7704 (gimplify_adjust_omp_clauses): Likewise.
7705 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
7706 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
7707 (gimplify_omp_target_update): Update handling of acc update and
7709 * omp-low.c (install_var_field): Remove unused parameter
7710 base_pointers_restrict.
7711 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
7712 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
7714 (omp_target_base_pointers_restrict_p): Delete.
7715 (scan_omp_target): Update call to scan_sharing_clauses.
7716 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
7718 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
7719 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
7720 (convert_local_omp_clauses): Likewise.
7721 * tree-pretty-print.c (dump_omp_clause): Likewise.
7722 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
7724 (omp_clause_code_name): Likewise.
7726 2018-06-20 Jakub Jelinek <jakub@redhat.com>
7729 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
7732 PR tree-optimization/86231
7733 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
7734 anti-range don't overwrite *vr0min before using it to compute *vr0max.
7736 2018-06-20 Tom de Vries <tdevries@suse.de>
7738 PR tree-optimization/86097
7739 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
7740 iv type if signedness of iv type is not the same as that of *nit.
7742 2018-06-20 Jakub Jelinek <jakub@redhat.com>
7744 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
7745 EDGE_EH edges, verify they are all EDGE_EH.
7747 2018-06-20 Maya Rashish <coypu@sdf.org>
7749 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
7751 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7753 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
7754 * config/aarch64/aarch64.c (xgene1_tunings): Add
7755 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
7756 (aarch64_mode_valid_for_sched_fusion_p):
7757 Allow 16-byte modes.
7758 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
7759 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
7761 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
7763 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
7764 * config/aarch64/iterators.md (VQ2): New mode iterator.
7766 2018-06-20 Martin Liska <mliska@suse.cz>
7768 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
7769 Change default ratio from 10 to 8.
7771 2018-06-20 Martin Liska <mliska@suse.cz>
7773 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7775 (bit_test_cluster::find_bit_tests): Likewise.
7776 (switch_decision_tree::analyze_switch_statement): Find clusters.
7777 * tree-switch-conversion.h (struct jump_table_cluster): Document
7780 2018-06-20 Martin Liska <mliska@suse.cz>
7782 * tree-switch-conversion.c (switch_conversion::collect):
7783 Record m_uniq property.
7784 (switch_conversion::expand): Bail out for special conditions.
7785 (group_cluster::~group_cluster): New.
7786 (group_cluster::group_cluster): Likewise.
7787 (group_cluster::dump): Likewise.
7788 (jump_table_cluster::emit): New.
7789 (switch_decision_tree::fix_phi_operands_for_edges): New.
7790 (struct case_node): Remove struct.
7791 (jump_table_cluster::can_be_handled): New.
7792 (case_values_threshold): Moved to header.
7793 (reset_out_edges_aux): Likewise.
7794 (jump_table_cluster::is_beneficial): New.
7795 (bit_test_cluster::can_be_handled): Likewise.
7796 (add_case_node): Remove.
7797 (bit_test_cluster::is_beneficial): New.
7798 (case_bit_test::cmp): New.
7799 (bit_test_cluster::emit): New.
7800 (expand_switch_as_decision_tree_p): Remove.
7801 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
7802 (fix_phi_operands_for_edge): Likewise.
7803 (switch_decision_tree::analyze_switch_statement): New.
7804 (compute_cases_per_edge): Move ...
7805 (switch_decision_tree::compute_cases_per_edge): ... here.
7806 (try_switch_expansion): Likewise.
7807 (switch_decision_tree::try_switch_expansion): Likewise.
7808 (record_phi_operand_mapping): Likewise.
7809 (switch_decision_tree::record_phi_operand_mapping): Likewise.
7810 (emit_case_decision_tree): Likewise.
7811 (switch_decision_tree::emit): Likewise.
7812 (balance_case_nodes): Likewise.
7813 (switch_decision_tree::balance_case_nodes): Likewise.
7814 (dump_case_nodes): Likewise.
7815 (switch_decision_tree::dump_case_nodes): Likewise.
7816 (emit_jump): Likewise.
7817 (switch_decision_tree::emit_jump): Likewise.
7818 (emit_cmp_and_jump_insns): Likewise.
7819 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
7820 (emit_case_nodes): Likewise.
7821 (switch_decision_tree::emit_case_nodes): Likewise.
7822 (conditional_probability): Remove.
7823 * tree-switch-conversion.h (enum cluster_type): New.
7825 (struct cluster): Likewise.
7826 (cluster::cluster): Likewise.
7827 (struct simple_cluster): Likewise.
7828 (simple_cluster::simple_cluster): Likewise.
7829 (struct group_cluster): Likewise.
7830 (struct jump_table_cluster): Likewise.
7831 (struct bit_test_cluster): Likewise.
7832 (struct min_cluster_item): Likewise.
7833 (struct case_tree_node): Likewise.
7834 (case_tree_node::case_tree_node): Likewise.
7835 (jump_table_cluster::case_values_threshold): Likewise.
7836 (struct case_bit_test): Likewise.
7837 (struct switch_decision_tree): Likewise.
7838 (struct switch_conversion): Likewise.
7839 (switch_decision_tree::reset_out_edges_aux): Likewise.
7841 2018-06-20 Martin Liska <mliska@suse.cz>
7843 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
7844 (hoist_edge_and_branch_if_true): Likewise.
7845 (expand_switch_using_bit_tests_p): Likewise.
7846 (struct case_bit_test): Likewise.
7847 (case_bit_test_cmp): Likewise.
7848 (emit_case_bit_tests): Likewise.
7849 (switch_conversion::switch_conversion): New class.
7850 (struct switch_conv_info): Remove old struct.
7851 (collect_switch_conv_info): More to ...
7852 (switch_conversion::collect): ... this.
7853 (check_range): Likewise.
7854 (switch_conversion::check_range): Likewise.
7855 (check_all_empty_except_final): Likewise.
7856 (switch_conversion::check_all_empty_except_final): Likewise.
7857 (check_final_bb): Likewise.
7858 (switch_conversion::check_final_bb): Likewise.
7859 (create_temp_arrays): Likewise.
7860 (switch_conversion::create_temp_arrays): Likewise.
7861 (free_temp_arrays): Likewise.
7862 (gather_default_values): Likewise.
7863 (switch_conversion::gather_default_values): Likewise.
7864 (build_constructors): Likewise.
7865 (switch_conversion::build_constructors): Likewise.
7866 (constructor_contains_same_values_p): Likewise.
7867 (switch_conversion::contains_same_values_p): Likewise.
7868 (array_value_type): Likewise.
7869 (switch_conversion::array_value_type): Likewise.
7870 (build_one_array): Likewise.
7871 (switch_conversion::build_one_array): Likewise.
7872 (build_arrays): Likewise.
7873 (switch_conversion::build_arrays): Likewise.
7874 (gen_def_assigns): Likewise.
7875 (switch_conversion::gen_def_assigns): Likewise.
7876 (prune_bbs): Likewise.
7877 (switch_conversion::prune_bbs): Likewise.
7878 (fix_phi_nodes): Likewise.
7879 (switch_conversion::fix_phi_nodes): Likewise.
7880 (gen_inbound_check): Likewise.
7881 (switch_conversion::gen_inbound_check): Likewise.
7882 (process_switch): Use the newly created class.
7883 (switch_conversion::expand): New.
7884 (switch_conversion::~switch_conversion): New.
7885 * tree-switch-conversion.h: New file.
7887 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7889 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
7890 tree-vect-patterns.c.
7891 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
7892 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
7893 (vect_recog_sad_pattern): Likewise.
7894 (vect_recog_widen_sum_pattern): Likewise.
7895 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
7896 (vect_recog_widen_shift_pattern): Remove the type_in argument.
7897 (vect_recog_rotate_pattern): Likewise.
7898 (vect_recog_mult_pattern): Likewise.
7899 (vect_recog_vector_vector_shift_pattern): Likewise.
7900 (vect_recog_divmod_pattern): Likewise.
7901 (vect_recog_mixed_size_cond_pattern): Likewise.
7902 (vect_recog_bool_pattern): Likewise.
7903 (vect_recog_mask_conversion_pattern): Likewise.
7904 (vect_try_gather_scatter_pattern): Likewise.
7905 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
7906 (vect_recog_over_widening_pattern): Likewise.
7907 (vect_recog_gather_scatter_pattern): Likewise.
7908 (vect_recog_func_ptr): Move from tree-vectorizer.h
7909 (vect_vect_recog_func_ptrs): Move further down the file.
7910 (vect_recog_func): Likewise. Remove the third argument.
7911 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
7912 (vect_pattern_recog_1): Expect the pattern function to do any
7913 necessary target tests. Also expect it to provide a vector type.
7914 Remove the type_in handling.
7916 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7918 * tree-vect-patterns.c (vect_pattern_detected): New function.
7919 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
7920 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
7921 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
7922 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
7923 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
7924 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
7925 (vect_recog_mask_conversion_pattern)
7926 (vect_try_gather_scatter_pattern): Likewise.
7928 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7930 * tree-vect-patterns.c (vect_get_internal_def): New function.
7931 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7932 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
7933 (search_type_for_mask_1): Use it.
7935 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7937 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
7938 redundant WIDEN_SUM_EXPR handling.
7939 (vect_recog_sad_pattern): Likewise.
7941 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7943 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
7944 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
7945 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
7946 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
7947 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
7949 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7951 * tree-vect-stmts.c (vectorizable_call): Make sure that we
7952 use the stmt_vec_info of the original bb statement for the
7953 new zero assignment, even if the call is part of a pattern.
7955 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7957 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
7958 that the sequence is attached to the original statement rather
7959 than the pattern statement.
7960 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
7961 PATTERN_DEF_SEQ from the original statement rather than
7962 the main pattern statement.
7963 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
7964 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
7965 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
7967 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7969 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
7970 definition statements before the early exit for statements that aren't
7972 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
7974 (vect_transform_loop): ...here. Process pattern definition
7975 statements without first checking whether the main pattern
7976 statement is live or relevant.
7978 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
7980 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
7981 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
7983 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
7985 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
7986 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
7987 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
7988 (expand_block_compare): Change select_block_compare_mode call.
7989 (expand_strncmp_align_check): Use new functions, fix comment.
7990 (emit_final_str_compare_gpr): New function.
7991 (expand_strn_compare): Refactor and clean up code.
7992 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
7994 2018-06-19 Tony Reix <tony.reix@atos.com>
7995 Damien Bergamini <damien.bergamini@atos.com>
7996 David Edelsohn <dje.gcc@gmail.com>
7998 * collect2.c (static_obj): New variable.
7999 (static_libs): New variable.
8000 (is_in_list): Uncomment declaration.
8001 (main): Track AIX libraries linked statically.
8002 (is_in_list): Uncomment definition.
8003 (scan_prog_file): Don't add AIX shared libraries initializer
8004 to constructor list if linking statically.
8006 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
8008 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
8010 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
8012 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
8014 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
8017 2018-06-19 Martin Liska <mliska@suse.cz>
8019 * config/i386/i386.c (ix86_can_inline_p): Do not use
8020 ipa_fn_summaries::get_create.
8021 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
8023 (devirtualization_time_bonus): Likewise.
8024 (ipcp_propagate_stage): Likewise.
8025 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
8026 (edge_set_predicate): Likewise.
8027 (evaluate_conditions_for_known_args): Likewise.
8028 (evaluate_properties_for_edge): Likewise.
8029 (ipa_call_summary::reset): Tranform to ...
8030 (ipa_call_summary::~ipa_call_summary): ... this.
8031 (ipa_fn_summary::reset): Transform to ...
8032 (ipa_fn_summary::~ipa_fn_summary): ... this.
8033 (ipa_fn_summary_t::remove): Rename to ...
8034 (ipa_fn_summary_t::remove_callees): ... this.
8035 (ipa_fn_summary_t::duplicate): Use placement new
8036 instead of memory copy.
8037 (ipa_call_summary_t::duplicate): Likewise.
8038 (ipa_call_summary_t::remove): Remove.
8039 (dump_ipa_call_summary): Change get_create to get.
8040 (ipa_dump_fn_summary): Dump only when summary exists.
8041 (analyze_function_body): Use symbol_summary::get instead
8043 (compute_fn_summary): Likewise.
8044 (estimate_edge_devirt_benefit): Likewise.
8045 (estimate_edge_size_and_time): Likewise.
8046 (inline_update_callee_summaries): Likewise.
8047 (remap_edge_change_prob): Likewise.
8048 (remap_edge_summaries): Likewise.
8049 (ipa_merge_fn_summary_after_inlining): Likewise.
8050 (write_ipa_call_summary): Likewise.
8051 (ipa_fn_summary_write): Likewise.
8052 (ipa_free_fn_summary): Likewise.
8053 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
8054 (struct ipa_call_summary): Likewise.
8055 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
8057 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
8058 (estimate_size_after_inlining): Likewise.
8059 (estimate_growth): Likewise.
8060 (growth_likely_positive): Likewise.
8061 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
8062 (inline_call): Likewise.
8063 * ipa-inline.c (caller_growth_limits): Likewise.
8064 (can_inline_edge_p): Likewise.
8065 (can_inline_edge_by_limits_p): Likewise.
8066 (compute_uninlined_call_time): Likewise.
8067 (compute_inlined_call_time): Likewise.
8068 (want_inline_small_function_p): Likewise.
8069 (edge_badness): Likewise.
8070 (update_caller_keys): Likewise.
8071 (update_callee_keys): Likewise.
8072 (inline_small_functions): Likewise.
8073 (inline_to_all_callers_1): Likewise.
8074 (dump_overall_stats): Likewise.
8075 (early_inline_small_functions): Likewise.
8076 (early_inliner): Likewise.
8077 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
8078 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
8079 * ipa-pure-const.c (malloc_candidate_p): Likewise.
8080 * ipa-split.c (execute_split_functions): Likewise.
8081 * symbol-summary.h: Likewise.
8082 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
8084 2018-06-19 Richard Biener <rguenther@suse.de>
8086 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
8087 (vectorize_loops): ... here. Fix dbgcnt handling.
8088 (try_vectorize_loop): Wrap try_vectorize_loop_1.
8090 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
8093 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
8094 ieee128 argument takes up only one (vector) register, not two (floating
8097 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
8099 * gimplify.c (gimplify_init_constructor): Really never clear for an
8100 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
8102 2018-06-19 Richard Biener <rguenther@suse.de>
8104 PR tree-optimization/86179
8105 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
8106 after failed recognition.
8108 2018-06-18 Martin Sebor <msebor@redhat.com>
8111 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
8112 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
8113 Handle integer subtraction.
8114 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
8115 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
8117 2018-06-18 David Malcolm <dmalcolm@redhat.com>
8119 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
8120 param from rtx to rtx_insn *.
8121 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
8123 (frv_ifcvt_modify_insn): Likwise.
8124 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
8125 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
8126 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
8127 as_a <rtx_insn *> cast to local "unprotected_region" once
8128 it's been established that it's not NULL or pc_rtx.
8129 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
8130 param "sethi" from rtx to rtx_insn *.
8131 (nds32_group_float_insns): Likewise for param "insn".
8132 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
8134 (vax_output_int_subtract): Likewise.
8135 * config/vax/vax.c (vax_output_int_add): Likewise for param
8137 (vax_output_int_subtract): Likewise.
8138 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
8139 (emit_pattern_after): Likewise for param "after".
8140 (emit_insn_after): Likewise.
8141 (emit_jump_insn_after): Likewise.
8142 (emit_call_insn_after): Likewise.
8143 (emit_debug_insn_after): Likewise.
8144 (emit_pattern_before): Likewise for param "before".
8145 (emit_insn_before): Likewise.
8146 (emit_jump_insn_before): Likewise.
8147 * final.c (get_insn_template): Likewise for param "insn", removing
8149 * output.h (get_insn_template): Likewise for 2nd param.
8150 * rtl.h (emit_insn_before): Likewise.
8151 (emit_jump_insn_before): Likewise.
8152 (emit_debug_insn_before_noloc): Likewise.
8153 (emit_insn_after): Likewise.
8154 (emit_jump_insn_after): Likewise.
8155 (emit_call_insn_after): Likewise.
8156 (emit_debug_insn_after): Likewise.
8157 (set_insn_deleted): Likewise for param.
8159 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
8162 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
8163 floating point modes, so that IFmode is numerically greater than
8164 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
8165 to declare the ordering. This prevents IFmode from being
8166 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
8167 machine. Include rs6000-modes.h to share the fractional values
8168 between genmodes* and the rest of the compiler.
8172 * config/rs6000/rs6000-modes.h: New file.
8173 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
8174 meaning of rs6000_long_double_size so that 126..128 selects an
8175 appropriate 128-bit floating point type.
8176 (rs6000_option_override_internal): Likewise.
8177 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
8178 (TARGET_LONG_DOUBLE_128): Change the meaning of
8179 rs6000_long_double_size so that 126..128 selects an appropriate
8180 128-bit floating point type.
8181 (LONG_DOUBLE_TYPE_SIZE): Update comment.
8182 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
8183 source and destination to match the standard usage.
8184 (truncifkf2): Likewise.
8185 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
8186 ISA 2.07 to use an explicit clobber, instead of passing in a
8188 (copysign<mode>3_soft): Likewise.
8190 2018-06-18 David Malcolm <dmalcolm@redhat.com>
8192 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
8193 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
8194 (vect_slp_analyze_instance_dependence): Likewise.
8195 (vect_enhance_data_refs_alignment): Likewise.
8196 (vect_analyze_data_refs_alignment): Likewise.
8197 (vect_slp_analyze_and_verify_instance_alignment
8198 (vect_analyze_data_ref_accesses): Likewise.
8199 (vect_prune_runtime_alias_test_list): Likewise.
8200 (vect_analyze_data_refs): Likewise.
8201 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
8202 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8203 (vect_analyze_scalar_cycles_1): Likewise.
8204 (vect_get_loop_niters): Likewise.
8205 (vect_analyze_loop_form_1): Likewise.
8206 (vect_update_vf_for_slp): Likewise.
8207 (vect_analyze_loop_operations): Likewise.
8208 (vect_analyze_loop): Likewise.
8209 (vectorizable_induction): Likewise.
8210 (vect_transform_loop): Likewise.
8211 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
8212 * tree-vect-slp.c (vect_analyze_slp): Likewise.
8213 (vect_make_slp_decision): Likewise.
8214 (vect_detect_hybrid_slp): Likewise.
8215 (vect_slp_analyze_operations): Likewise.
8216 (vect_slp_bb): Likewise.
8217 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
8218 (vectorizable_bswap): Likewise.
8219 (vectorizable_call): Likewise.
8220 (vectorizable_simd_clone_call): Likewise.
8221 (vectorizable_conversion): Likewise.
8222 (vectorizable_assignment): Likewise.
8223 (vectorizable_shift): Likewise.
8224 (vectorizable_operation): Likewise.
8225 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
8227 2018-06-18 Martin Sebor <msebor@redhat.com>
8229 PR tree-optimization/81384
8230 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
8231 * builtins.c (expand_builtin_strnlen): New function.
8232 (expand_builtin): Call it.
8233 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
8234 * builtins.def (BUILT_IN_STRNLEN): New.
8235 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
8236 Warn for bounds in excess of maximum object size.
8237 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
8238 single-value ranges. Handle strnlen.
8239 (handle_builtin_strlen): Handle strnlen.
8240 (strlen_check_and_optimize_stmt): Same.
8241 * doc/extend.texi (Other Builtins): Document strnlen.
8243 2018-06-18 Maya Rashish <coypu@sdf.org>
8245 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
8246 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
8247 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
8249 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
8251 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
8253 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8255 * tree.c (escaped_string::escape): Replace cast to char * by
8256 const_cast<char *> (unescaped).
8258 2018-06-18 Nick Clifton <nickc@redhat.com>
8261 * tree.c (escaped_string): New class. Converts an unescaped
8262 string into its escaped equivalent.
8263 (warn_deprecated_use): Use the new class to convert the
8264 deprecation message, if present.
8265 (test_escaped_strings): New self test.
8266 (test_c_tests): Add test_escaped_strings.
8267 * doc/extend.texi (deprecated): Add a note that the
8268 deprecation message is affected by the -fmessage-length
8269 option, and that control characters will be escaped.
8270 (#pragma GCC error): Document this pragma.
8271 (#pragma GCC warning): Likewise.
8272 * doc/invoke.texi (-fmessage-length): Document this option's
8273 effect on the #warning and #error preprocessor directives and
8274 the deprecated attribute.
8276 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
8278 * tree.c (decl_value_expr_lookup): Revert latest change.
8279 (decl_value_expr_insert): Likewise.
8281 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
8283 * gimplify.c (nonlocal_vlas): Delete.
8284 (nonlocal_vla_vars): Likewise.
8285 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
8287 (gimplify_body): Do not create and destroy nonlocal_vlas.
8288 * tree-nested.c: Include diagnostic.h.
8289 (use_pointer_in_frame): Tweak.
8290 (lookup_field_for_decl): Add assertion and declare the transformation.
8291 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
8292 internal error when the reference is in a wrong context. Do not
8293 create a debug decl by default.
8294 (note_nonlocal_block_vlas): Delete.
8295 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
8296 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
8297 create a debug decl by default.
8298 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
8299 call is in a wrong context.
8300 (fixup_vla_decls): New function.
8301 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
8302 debug variables were created.
8303 * tree.c (decl_value_expr_lookup): Add checking assertion.
8304 (decl_value_expr_insert): Likewise.
8306 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
8309 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
8310 * tree-scalar-evolution.c (interpret_expr): Likewise.
8311 (expression_expensive_p): Likewise.
8312 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
8313 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
8314 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
8315 (ssa_defined_by_minus_one_stmt_p): New.
8317 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
8320 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
8321 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
8322 * dojump.c (do_jump): Likewise.
8323 * expr.c (expand_expr_real_2): Check operand type's sign.
8324 * fold-const.c (const_unop): Handle ABSU_EXPR.
8325 (fold_abs_const): Likewise.
8326 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
8327 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
8328 (strip_sign_op_1): Likesise.
8329 * match.pd: Add new pattern to generate ABSU_EXPR.
8330 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
8331 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
8332 * tree-eh.c (operation_could_trap_helper_p): Likewise.
8333 * tree-inline.c (estimate_operator_cost): Likewise.
8334 * tree-pretty-print.c (dump_generic_node): Likewise.
8335 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
8336 * tree.def (ABSU_EXPR): New.
8338 2018-06-16 Jakub Jelinek <jakub@redhat.com>
8341 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
8342 documented as preserved for backward compatibility only.
8343 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
8345 PR rtl-optimization/86108
8346 * bb-reorder.c (create_forwarder_block): Renamed to ...
8347 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
8348 jump from new landing pad to the second part.
8349 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
8352 2018-06-15 Jakub Jelinek <jakub@redhat.com>
8355 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
8356 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
8357 Only call store_expr for halves if the mode is the same.
8360 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
8361 Fix up comment formatting.
8363 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
8365 * typed-splay-tree.h (typed_splay_tree::remove): New function.
8366 (typed_splay_tree::closure,
8367 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
8368 (typed_splay_tree::typed_splay_tree,
8369 typed_splay_tree::operator =): Declared private.
8370 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
8371 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
8372 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
8373 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
8374 typed_splay_tree::splay_tree_splay,
8375 typed_splay_tree::splay_tree_foreach_helper,
8376 typed_splay_tree::splay_tree_insert,
8377 typed_splay_tree::splay_tree_remove,
8378 typed_splay_tree::splay_tree_lookup,
8379 typed_splay_tree::splay_tree_predecessor,
8380 typed_splay_tree::splay_tree_successor,
8381 typed_splay_tree::splay_tree_min,
8382 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
8383 (typed_splay_tree::root, typed_splay_tree::comp,
8384 typed_splay_tree::delete_key,
8385 typed_splay_tree::delete_value): New data members.
8386 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
8387 typed_splay_tree::remove.
8389 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
8391 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
8392 -mginv and -mno-ginv to the assembler.
8393 * config/mips/mips.opt (-mcrc): New option.
8395 * doc/invoke.text (-mcrc): Document.
8398 2018-06-15 Nick Clifton <nickc@redhat.com>
8401 * tree.c (escaped_string): New class. Converts an unescaped
8402 string into its escaped equivalent.
8403 (warn_deprecated_use): Use the new class to convert the
8404 deprecation message, if present.
8405 (test_escaped_strings): New self test.
8406 (test_c_tests): Add test_escaped_strings.
8407 * doc/extend.texi (deprecated): Add a note that the
8408 deprecation message is affected by the -fmessage-length
8409 option, and that control characters will be escaped.
8410 (#pragma GCC error): Document this pragma.
8411 (#pragma GCC warning): Likewise.
8412 * doc/invoke.texi (-fmessage-length): Document this option's
8413 effect on the #warning and #error preprocessor directives and
8414 the deprecated attribute.
8416 2018-06-15 Richard Biener <rguenther@suse.de>
8418 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
8419 here, also noting vector size used.
8420 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
8421 size used in MSG_OPTIMIZED_LOCATIONS dump.
8422 (pass_slp_vectorize::execute): Adjust.
8424 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
8427 * config/arc/arc.c (arc_return_address_register): Fix
8430 2018-06-15 Richard Biener <rguenther@suse.de>
8433 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
8434 leave useless conversion stripping to force_gimple_operand_gsi.
8435 (gimplify_build2): Likewise.
8436 (gimplify_build1): Likewise.
8438 2018-06-15 Richard Biener <rguenther@suse.de>
8441 * tree-cfg.c (move_stmt_op): unshare invariant addresses
8442 before adjusting their block.
8444 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
8446 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
8447 multilibs for *-*-rtems*.
8448 * config/riscv/t-rtems: New file.
8450 2018-06-14 Jakub Jelinek <jakub@redhat.com>
8453 * match.pd ((A +- CST1) +- CST2): Punt if last resort
8454 unsigned_type_for returns NULL.
8457 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
8458 subregs of multi-word pseudos unless the float mode has word size.
8460 2018-06-14 Richard Biener <rguenther@suse.de>
8463 * tree-vect-generic.c (build_word_mode_vector_type): Remove
8464 duplicate and harmful type_hash_canon.
8465 * tree.c (type_hash_canon): Assert we didn't find ourselves.
8467 2018-06-14 Richard Biener <rguenther@suse.de>
8470 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
8473 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
8475 * config/rtems.h (STDINT_LONG32): Define.
8477 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
8478 Prachi Godbole <prachi.godbole@imgtec.com>
8480 * config/mips/mips-cpus.def: Define P6600.
8481 * config/mips/mips-tables.opt: Regenerate.
8482 * config/mips/mips.c (mips_ucbranch_type): New enum.
8483 (mips_rtx_cost_data): Add support for P6600.
8484 (mips_issue_rate): Likewise.
8485 (mips_multipass_dfa_lookahead): Likewise.
8486 (mips_avoid_hazard): Likewise.
8487 (mips_reorg_process_insns): Likewise.
8488 (mips_classify_branch_p6600): New function.
8489 * config/mips/mips.h (TUNE_P6600): New define.
8490 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
8491 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
8492 * config/mips/mips.md: Include p6600.md.
8493 (processor): Add p6600.
8494 * config/mips/p6600.md: New file.
8495 * doc/invoke.texi: Add p6600 to supported architectures.
8497 2018-06-13 Martin Sebor <msebor@redhat.com>
8499 PR tree-optimization/86114
8500 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
8502 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
8504 2018-06-13 Richard Biener <rguenther@suse.de>
8506 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
8507 Properly set vector type of the intermediate stmt.
8508 * tree-vect-stmts.c (vectorizable_operation): The destination
8509 var always has vectype_out type.
8511 2018-06-13 Jeff Law <law@redhat.com>
8513 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
8514 integer 0 for argument to print_rtl_with_bb.
8515 (rl78_reorg): Likewise.
8517 2018-06-13 David Malcolm <dmalcolm@redhat.com>
8519 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
8520 from rtx to rtx_insn *.
8521 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
8523 (add_sched_insns_for_speculation): Likewise for local "target",
8524 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
8525 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
8526 from rtx_insn ** to rtx_code_label **.
8527 (reorg_emit_nops): Likewise.
8528 (c6x_reorg): Likewise for local "call_labels".
8529 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
8531 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
8532 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
8533 the loops over LABEL_REFS.
8534 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
8536 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
8537 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
8538 (split_branches): Strengthen local "olabel" from rtx to
8539 rtx_insn *, adding a safe_as_a cast.
8540 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
8542 (add_insn_after): Likewise for first two params.
8543 (add_insn_before): Likewise.
8544 (remove_insn): Likewise for param.
8545 (emit_pattern_before_noloc): Likewise for second and third params.
8546 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
8547 (emit_call_insn_before_noloc): Likewise.
8548 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
8550 (emit_barrier_before): Likewise.
8551 (emit_label_before): Strengthen "label" param from "rtx" to
8552 "rtx_code_label *". Strengthen "before" param from "rtx" to
8554 (emit_insn_after_1): Strengthen "after" param from "rtx" to
8556 (emit_pattern_after_noloc): Likewise.
8557 (emit_insn_after_noloc): Likewise.
8558 (emit_jump_insn_after_noloc): Likewise.
8559 (emit_call_insn_after_noloc): Likewise.
8560 (emit_debug_insn_after_noloc): Likewise.
8561 (emit_barrier_after): Likewise.
8562 (emit_label_after): Likewise for both params.
8563 (emit_pattern_after_setloc): Likewise for "after" param. Convert
8564 "loc" param from "int" to "location_t".
8565 (emit_insn_after_setloc): Likewise.
8566 (emit_jump_insn_after_setloc): Likewise.
8567 (emit_call_insn_after_setloc): Likewise.
8568 (emit_debug_insn_after_setloc): Likewise.
8569 (emit_pattern_before_setloc): Likewise for "before" param. Convert
8570 "loc" param from "int" to "location_t".
8571 (emit_pattern_before): Convert NULL_RTX to NULL.
8572 (emit_insn_before_setloc): Convert "loc" param from "int" to
8574 (emit_jump_insn_before_setloc): Likewise.
8575 (emit_call_insn_before_setloc): Likewise.
8576 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
8577 rtx_insn *. Convert "loc" param from "int" to "location_t".
8578 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
8579 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
8580 Convert 3rd param from "int" to "location_t".
8581 (emit_barrier_before, emit_barrier_after, next_real_insn):
8582 Strengthen param from rtx to rtx_insn *.
8583 (emit_label_before): Strengthen 1st param from "rtx" to
8584 "rtx_code_label *". Strengthen 2nd param from "rtx" to
8586 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
8587 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
8588 Strengthen 2nd param from "rtx" to "rtx_insn *".
8589 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
8590 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
8591 Likewise. Convert 3rd param from "int" to "location_t".
8592 (emit_label_after): Strengthen 1st param from "rtx" to
8594 (next_real_insn, remove_insn): Strengthen param from "rtx" to
8596 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
8597 from "rtx" to "rtx_insn *".
8599 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
8601 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
8602 bodies streamed in with -Q.
8603 * dumpfile.c (dump_files): Add lto-stream-out dump file.
8604 * dumpfile.h (tree_dump_index): Add lto_stream_out.
8605 * gimple-streamer-out.c: Include gimple-pretty-print.h
8606 (output_bb): Dump stmts streamed.
8607 * lto-section-out.c: Include print-tree.h
8608 (lto_begin_section): Dump sections created.
8609 (lto_output_decl_index): Dump decl encoded.
8610 * lto-streamer-out.c: Include print-tree.h
8611 (create_output_block): Dump output block created.
8612 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
8613 (output_function): Dump function output.
8614 (output_constructor): Dump constructor streamed.
8615 (write_global_stream): Output indexes encoded.
8616 (produce_asm_for_decls): Dump streams encoded.
8617 * lto-streamer.c (streamer_dump_file): New global var.
8618 * lto-streamer.h (streamer_dump_file): Declare.
8619 * passes.c (ipa_write_summaries): Initialize streamer dump.
8620 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
8623 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
8626 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
8627 offsets for register save directives. Emit a second batch of save
8628 directives, if need be, when the function accesses prior frames.
8630 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8632 * config/arc/fpu.md (fmasf4): Force operand to register.
8633 (fnmasf4): Likewise.
8635 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8637 * config/arc/arc-protos.h (arc_pad_return): Remove.
8638 * config/arc/arc.c (machine_function): Remove force_short_suffix
8640 (arc_print_operand): Adjust printing of '&'.
8641 (arc_verify_short): Remove conditional printing of short suffix.
8642 (arc_final_prescan_insn): Remove reference to size_reason.
8643 (pad_return): New function.
8644 (arc_reorg): Call pad_return.
8645 (arc_pad_return): Remove.
8646 (arc_init_machine_status): Remove reference to force_short_suffix.
8647 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
8648 (attr length): When attribute iscompact is true force to 2
8649 regardless; in the case of maybe check if we want to force the
8650 instruction to have 4 bytes length.
8651 (nopv): Change it to generate 4 byte long nop as well.
8652 (blockage): New pattern.
8653 (simple_return): Remove call to arc_pad_return.
8654 (p_return_i): Likewise.
8656 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8658 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
8660 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8662 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
8665 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8667 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
8668 for ARC700 and ARCv2.
8670 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
8673 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
8674 operands[2] instead of operands[1].
8677 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
8679 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
8680 case, check whether the outer register overlaps an unallocatable
8681 register, not just whether it fits the required class.
8683 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
8685 * poly-int.h (can_div_trunc_p): Add new overload in which all values
8687 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
8688 (memrefs_conflict_p): Likewise.
8689 (init_alias_analysis): Likewise.
8690 * cfgexpand.c (expand_debug_expr): Likewise.
8691 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
8692 * cse.c (fold_rtx): Likewise.
8693 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
8694 * expr.c (emit_block_move_hints): Likewise.
8695 (clear_storage_hints, push_block, emit_push_insn): Likewise.
8696 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
8697 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
8698 (emit_group_store): Likewise.
8699 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
8700 to read the PRE/POST_MODIFY increment.
8701 * calls.c (store_one_arg): Use strip_offset.
8702 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
8704 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
8706 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
8707 (simplify_binary_operation_1): Extend CONST_INT handling to
8709 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
8710 than a HOST_WIDE_INT.
8711 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
8713 (adjust_mems, add_stores): Update accodingly.
8714 (vt_canonicalize_addr): Track polynomial offsets.
8715 (emit_note_insn_var_location): Likewise.
8716 (vt_add_function_parameter): Likewise.
8717 (vt_initialize): Likewise.
8719 2018-06-12 Jeff Law <law@redhat.com>
8721 * config.gcc (alpha*-*-freebsd*): Remove.
8722 * config/alpha/freebsd.h: Remove.
8724 2018-06-12 David Malcolm <dmalcolm@redhat.com>
8727 * spellcheck-tree.c (levenshtein_distance): Rename to...
8728 (get_edit_distance): ...this, and update for underlying renaming.
8729 * spellcheck-tree.h (levenshtein_distance): Rename to...
8730 (get_edit_distance): ...this.
8731 * spellcheck.c (levenshtein_distance): Rename to...
8732 (get_edit_distance): ...this. Convert from Levenshtein distance
8733 to Damerau-Levenshtein distance by supporting transpositions of
8734 adjacent characters. Rename "v1" to "v_next" and "v0" to
8736 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
8737 (selftest::test_edit_distance_unit_test_oneway): ...this, and
8738 update for underlying renaming.
8739 (selftest::levenshtein_distance_unit_test): Rename to...
8740 (selftest::test_get_edit_distance_unit): ...this, and update for
8741 underlying renaming.
8742 (selftest::test_find_closest_string): Add example from PR 69968
8743 where transposition helps
8744 (selftest::test_metric_conditions): Update for renaming.
8745 (selftest::test_metric_conditions): Likewise.
8746 (selftest::spellcheck_c_tests): Likewise.
8747 * spellcheck.h (levenshtein_distance): Rename both overloads to...
8748 (get_edit_distance): ...this.
8749 (best_match::consider): Update for renaming.
8751 2018-06-12 Martin Sebor <msebor@redhat.com>
8753 PR tree-optimization/85259
8754 * builtins.c (compute_objsize): Handle constant offsets.
8755 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
8756 true iff a warning has been issued.
8757 * gimple.h (gimple_nonartificial_location): New function.
8758 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
8759 gimple_nonartificial_location and handle -Wno-system-headers.
8760 (handle_builtin_stxncpy): Same.
8762 2018-06-12 Martin Sebor <msebor@redhat.com>
8765 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
8767 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
8769 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
8770 BUILTIN_VEC_XST entries for pointer to double and long long.
8772 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
8775 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
8777 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
8780 2018-06-12 Martin Liska <mliska@suse.cz>
8782 * doc/options.texi: Document IntegerRange.
8784 2018-06-12 Martin Liska <mliska@suse.cz>
8786 * config/i386/i386.opt: Make MPX-related options as Deprecated.
8787 * opt-functions.awk: Handle Deprecated flag.
8788 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
8790 (read_cmdline_option): Report warning for a deprecated option.
8791 * opts.h (struct cl_option): Add new field cl_deprecated.
8792 (CL_ERR_DEPRECATED): New.
8794 2018-06-12 Martin Liska <mliska@suse.cz>
8796 * doc/options.texi: Document Deprecated option flag.
8798 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8800 * config/arc/arc-arch.h (arc_extras): New enum.
8801 (arc_cpu_t):Add field extra.
8802 (arc_cpu_types): Consider the extras.
8803 * config/arc/arc-cpus.def: Add extras info.
8804 * config/arc/arc-opts.h (processor_type): Consider extra field.
8805 * config/arc/arc.c (arc_override_options): Handle extra field.
8807 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8809 * config/arc/arc-arch.h: Update ARC_OPTX macro.
8810 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
8812 * config/arc/arc.c (arc_init): Update pic warning.
8813 (irq_range): Update irq range parsing warnings.
8814 (arc_override_options): Update various warning messages.
8815 (arc_handle_aux_attribute): Likewise.
8817 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
8819 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
8821 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8823 * doc/sourcebuild.texi: Document usage of line number 0 in verify
8824 compiler messages directives.
8826 2018-06-12 Matthew Fortune <mfortune@gmail.com>
8828 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
8829 * config/mips/mips-tables.opt: Regenerate.
8830 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
8832 * doc/invoke.texi: Document -march=i6500.
8834 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
8836 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
8837 (i6400_gpmul): Add cpu_unit.
8838 (i6400_gpdiv): Likewise.
8839 (i6400_msa_add_d): Update reservations.
8840 (i6400_msa_int_add) Likewise.
8841 (i6400_msa_short_logic3) Likewise.
8842 (i6400_msa_short_logic2) Likewise.
8843 (i6400_msa_short_logic) Likewise.
8844 (i6400_msa_move) Likewise.
8845 (i6400_msa_cmp) Likewise.
8846 (i6400_msa_short_float2) Likewise.
8847 (i6400_msa_div_d) Likewise.
8848 (i6400_msa_long_logic1) Likewise.
8849 (i6400_msa_long_logic2) Likewise.
8850 (i6400_msa_mult) Likewise.
8851 (i6400_msa_long_float2) Likewise.
8852 (i6400_msa_long_float4) Likewise.
8853 (i6400_msa_long_float5) Likewise.
8854 (i6400_msa_long_float8) Likewise.
8855 (i6400_fpu_fadd): Include frint type.
8856 (i6400_fpu_store): New define_insn_reservation.
8857 (i6400_fpu_load): Likewise.
8858 (i6400_fpu_move): Likewise.
8859 (i6400_fpu_fcmp): Likewise.
8860 (i6400_fpu_fmadd): Likewise.
8861 (i6400_int_mult): Include imul3nc type and update reservation.
8862 (i6400_int_div): Include idiv3 type and update reservation.
8863 (i6400_int_load): Update to check type not move_type.
8864 (i6400_int_store): Likewise.
8865 (i6400_int_prefetch): Set zero latency.
8867 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
8869 * gcc.c: Document new %@{...} sequence.
8870 (LINK_COMMAND_SPEC): Use it for the -L switches.
8871 (cpp_unique_options): Use it for the -I switches.
8872 (at_file_argbuf): New global variable.
8873 (in_at_file): Likewise.
8874 (alloc_args): Create at_file_argbuf.
8875 (clear_args): Truncate at_file_argbuf.
8876 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
8877 (open_at_file): New function.
8878 (close_at_file): Likewise.
8879 (create_at_file): Delete.
8880 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
8883 (validate_switches_from_spec): Deal with %@{...} sequence.
8884 (validate_switches): Likewise.
8885 (driver::finalize): Call clear_args.
8887 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
8889 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
8891 2018-06-11 Martin Sebor <msebor@redhat.com>
8893 * doc/invoke.texi (-Wall): List -Wc++17-compat.
8894 (Wno-class-memaccess): Add @opindex.
8895 (Wno-templates, Wno-multiple-inheritance): Same.
8896 (Wno-virtual-inheritance, Wno-namespaces): Same.
8897 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
8898 (Wno-format-overflow, Wno-format-truncation): Same.
8899 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
8900 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
8901 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
8902 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
8903 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
8904 (Wno-misspelled-isr): Same.
8906 2018-06-11 Martin Sebor <msebor@redhat.com>
8908 * PR tree-optimization/86083
8909 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
8911 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
8913 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
8915 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
8918 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
8919 on the correct operand.
8920 (*movdi_internal64): Ditto.
8922 2018-06-11 Martin Liska <mliska@suse.cz>
8924 PR tree-optimization/86089
8925 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
8927 2018-06-11 Julia Koval <julia.koval@intel.com>
8929 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
8930 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
8931 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
8933 2018-06-11 Olivier Hainque <hainque@adacore.com>
8935 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
8936 for Ada with strict dwarf2.
8938 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
8941 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
8944 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
8946 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
8948 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
8950 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
8953 2018-06-08 Martin Liska <mliska@suse.cz>
8955 * tree-cfg.h (debug_function): Fix argument type to match
8958 2018-06-08 Martin Liska <mliska@suse.cz>
8960 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
8961 Remove usage of MPX-related (and removed) fields.
8962 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
8964 2018-06-08 David Malcolm <dmalcolm@redhat.com>
8966 * cfg.c (debug): Use TDF_NONE rather than 0.
8967 * cfghooks.c (debug): Likewise.
8968 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
8969 (struct dump_option_value_info): Convert to...
8970 (struct kv_pair): ...this template type.
8971 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
8973 (optinfo_verbosity_options): Likewise.
8974 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
8976 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
8977 than int for "optgroup_flags" param.
8978 (dump_generic_expr_loc): Use dump_flags_t rather than int for
8980 (dump_dec): Likewise.
8981 (dump_finish): Use TDF_NONE rather than 0.
8982 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
8983 rather than int for "optgroup_flags" param. Use TDF_NONE rather
8984 than 0. Update for change to option_ptr.
8985 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
8986 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
8987 0. Update for changes to optinfo_verbosity_options and
8989 (opt_info_switch_p): Convert optgroup_flags from int to
8991 (dump_basic_block): Use dump_flags_t rather than int
8992 for "dump_kind" param.
8993 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
8994 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
8995 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
8996 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
8997 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
8998 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
8999 TDF_NONE): Convert from macros to...
9000 (enum dump_flag): ...this new enum.
9001 (dump_flags_t): Update to use enum.
9002 (operator|, operator&, operator~, operator|=, operator&=):
9003 Implement for dump_flags_t.
9004 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
9005 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
9006 Convert from macros to...
9007 (enum optgroup_flag): ...this new enum.
9008 (optgroup_flags_t): New typedef.
9009 (operator|, operator|=): Implement for optgroup_flags_t.
9010 (struct dump_file_info): Convert field "alt_flags" to
9011 dump_flags_t. Convert field "optgroup_flags" to
9013 (dump_basic_block): Use dump_flags_t rather than int for param.
9014 (dump_generic_expr_loc): Likewise.
9015 (dump_dec): Likewise.
9016 (dump_register): Convert param "optgroup_flags" to
9018 (opt_info_enable_passes): Likewise.
9019 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
9021 * gimple-pretty-print.c (debug): Likewise.
9022 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
9023 (merged_store_group::apply_stores): Likewise.
9024 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
9025 * gimple.c (verify_gimple_pp): Likewise.
9026 * graphite-poly.c (print_pbb_body): Likewise.
9027 * passes.c (pass_manager::register_one_dump_file): Convert
9028 local "optgroup_flags" to optgroup_flags_t.
9029 * print-tree.c (print_node): Use TDF_NONE rather than 0.
9031 (debug_body): Likewise.
9032 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
9033 to optgroup_flags_t.
9034 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
9036 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
9037 (convert_mult_to_fma): Likewise.
9038 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
9039 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
9040 * tree-vect-data-refs.c (dump_lower_bound): Convert param
9041 "dump_kind" to dump_flags_t.
9043 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
9045 * config/rs6000/rs6000.c (min, max): Delete.
9047 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
9049 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
9050 -mabi=spe and -mabi=no-spe.
9052 2018-06-08 Martin Liska <mliska@suse.cz>
9054 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
9055 where we expect an existing summary.
9057 2018-06-08 Martin Liska <mliska@suse.cz>
9059 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
9060 * ipa-inline.h (estimate_edge_growth): Likewise.
9062 2018-06-08 Martin Liska <mliska@suse.cz>
9064 * cgraph.c (function_version_hasher::hash): Use
9065 cgraph_node::get_uid ().
9066 (function_version_hasher::equal):
9067 * cgraph.h (cgraph_node::get_uid): New method.
9068 * ipa-inline.c (update_caller_keys): Use
9069 cgraph_node::get_uid ().
9070 (update_callee_keys): Likewise.
9071 * ipa-utils.c (searchc): Likewise.
9072 (ipa_reduced_postorder): Likewise.
9073 * lto-cgraph.c (input_node): Likewise.
9074 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
9075 * symbol-summary.h (symtab_insertion): Likewise.
9076 (symtab_removal): Likewise.
9077 (symtab_duplication): Likewise.
9078 * tree-pretty-print.c (dump_function_header): Likewise.
9079 * tree-sra.c (convert_callers_for_node): Likewise.
9081 2018-06-08 Martin Liska <mliska@suse.cz>
9083 * cgraph.c (symbol_table::create_edge): Always assign a new
9085 (symbol_table::free_edge): Do not recycle numbers.
9086 * cgraph.h (cgraph_edge::get): New method.
9087 * symbol-summary.h (symtab_removal): Use it.
9088 (symtab_duplication): Likewise.
9089 (call_summary::hashable_uid): Remove.
9091 2018-06-08 Martin Liska <mliska@suse.cz>
9093 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
9094 (initialize_growth_caches): Remove.
9095 (free_growth_caches): Likewise.
9096 (do_estimate_edge_time): Use edge_growth_cache.
9097 (do_estimate_edge_size): Likewise.
9098 (do_estimate_edge_hints): Likewise.
9099 * ipa-inline.c (reset_edge_caches): Likewise.
9100 (recursive_inlining): Likewise.
9101 (inline_small_functions): Likewise.
9102 * ipa-inline.h (initialize_growth_caches): Remove.
9103 (estimate_edge_size): Likewise.
9104 (estimate_edge_time): Likewise.
9105 (estimate_edge_hints): Likewise.
9106 (reset_edge_growth_cache): Likewise.
9107 * symbol-summary.h (call_summary::remove): New method.
9109 2018-06-08 Martin Liska <mliska@suse.cz>
9111 * ipa-cp.c (class edge_clone_summary): New summary.
9112 (grow_edge_clone_vectors): Remove.
9113 (ipcp_edge_duplication_hook): Remove.
9114 (class edge_clone_summary_t): New call_summary class.
9115 (ipcp_edge_removal_hook): Remove.
9116 (edge_clone_summary_t::duplicate): New function.
9117 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
9118 (create_specialized_node): Likewise.
9119 (ipcp_driver): Initialize edge_clone_summaries and do not
9122 2018-06-08 Martin Liska <mliska@suse.cz>
9124 * symbol-summary.h (get): New function.
9125 (call_summary::m_initialize_when_cloning): New class member.
9127 2018-06-08 Martin Liska <mliska@suse.cz>
9129 * cgraph.c (cgraph_node::remove): Do not recycle uid.
9130 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
9131 (symbol_table::allocate_cgraph_symbol): Do not set uid.
9132 * passes.c (uid_hash_t): Record removed_nodes by their uids.
9133 (remove_cgraph_node_from_order): Use the removed_nodes set.
9134 (do_per_function_toporder): Likwise.
9135 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
9136 instead of summary_uid.
9137 (symtab_removal): Likewise.
9138 (symtab_duplication): Likewise.
9140 2018-06-08 Martin Liska <mliska@suse.cz>
9142 * ipa-cp.c (ipcp_store_bits_results): Use
9143 ipcp_transformation_sum.
9144 (ipcp_store_vr_results): Likewise.
9145 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
9147 (ipcp_transformation_initialize): ... this.
9148 (ipa_set_node_agg_value_chain):
9149 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
9150 (write_ipcp_transformation_info): Likewise.
9151 (read_ipcp_transformation_info): Likewise.
9152 (ipcp_update_bits): Likewise.
9153 (ipcp_update_vr): Likewise.
9154 (ipcp_transform_function): Likewise.
9155 * ipa-prop.h: Rename ipcp_transformation_summary to
9156 ipcp_transformation.
9157 (class ipcp_transformation_t): New function summary.
9158 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
9159 (ipa_get_agg_replacements_for_node): Likewise.
9161 2018-06-08 Martin Liska <mliska@suse.cz>
9163 * ipa-pure-const.c (struct funct_state_d): Do it class instead
9165 (class funct_state_summary_t): New function_summary class.
9166 (has_function_state): Remove.
9167 (get_function_state): Likewise.
9168 (set_function_state): Likewise.
9169 (add_new_function): Likewise.
9170 (funct_state_summary_t::insert): New function.
9171 (duplicate_node_data): Remove.
9172 (remove_node_data): Remove.
9173 (funct_state_summary_t::duplicate): New function.
9174 (register_hooks): Create new funct_state_summaries.
9175 (pure_const_generate_summary): Use it.
9176 (pure_const_write_summary): Likewise.
9177 (pure_const_read_summary): Likewise.
9178 (propagate_pure_const): Likewise.
9179 (propagate_nothrow): Likewise.
9180 (dump_malloc_lattice): Likewise.
9181 (propagate_malloc): Likewise.
9182 (execute): Do not register hooks, just remove summary
9184 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
9187 2018-06-08 Martin Liska <mliska@suse.cz>
9189 * ipa-reference.c (remove_node_data): Remove.
9190 (duplicate_node_data): Likewise.
9191 (class ipa_ref_var_info_summary_t): New class.
9192 (class ipa_ref_opt_summary_t): Likewise.
9193 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
9194 (get_reference_optimization_summary): Use
9195 ipa_ref_opt_sum_summaries.
9196 (set_reference_vars_info): Remove.
9197 (set_reference_optimization_summary): Likewise.
9198 (ipa_init): Create summaries.
9199 (init_function_info): Use function summary.
9200 (ipa_ref_opt_summary_t::duplicate): New function.
9201 (ipa_ref_opt_summary_t::remove): New function.
9202 (get_read_write_all_from_node): Fix GNU coding style.
9203 (propagate): Use function summary.
9204 (write_node_summary_p): Fix GNU coding style.
9205 (stream_out_bitmap): Likewise.
9206 (ipa_reference_read_optimization_summary): Use function summary.
9207 (ipa_reference_c_finalize): Do not release hooks.
9209 2018-06-08 Martin Liska <mliska@suse.cz>
9211 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
9212 (analyze_function_body): Extract multiple calls of get_create.
9213 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
9214 * ipa-inline.c (recursive_inlining): Use ::get method.
9215 * ipa-inline.h (estimate_edge_growth): Likewise.
9217 2018-06-08 Martin Liska <mliska@suse.cz>
9219 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
9221 (hsa_function_summary::hsa_function_summary): Use the new enum
9223 (hsa_gpu_implementation_p): Use hsa_summaries::get.
9224 * hsa-gen.c (hsa_get_host_function): Likewise.
9225 (get_brig_function_name): Likewise.
9226 * ipa-hsa.c (process_hsa_functions): Likewise.
9227 (ipa_hsa_write_summary): Likewise.
9228 * symbol-summary.h (symtab_duplication): Use ::get function/
9229 (get): New function.
9231 2018-06-08 Martin Liska <mliska@suse.cz>
9233 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
9235 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
9236 (hsa_register_kernel): Likewise.
9237 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
9238 * hsa-gen.c (hsa_get_host_function): Likewise.
9239 (get_brig_function_name): Likewise.
9240 (generate_hsa): Likewise.
9241 (pass_gen_hsail::execute): Likewise.
9242 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
9243 (devirtualization_time_bonus): Likewise.
9244 (ipcp_propagate_stage): Likewise.
9245 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
9246 (edge_set_predicate): Likewise.
9247 (evaluate_conditions_for_known_args): Likewise.
9248 (evaluate_properties_for_edge): Likewise.
9249 (ipa_fn_summary::reset): Likewise.
9250 (ipa_fn_summary_t::duplicate): Likewise.
9251 (dump_ipa_call_summary): Likewise.
9252 (ipa_dump_fn_summary): Likewise.
9253 (analyze_function_body): Likewise.
9254 (compute_fn_summary): Likewise.
9255 (estimate_edge_devirt_benefit): Likewise.
9256 (estimate_edge_size_and_time): Likewise.
9257 (estimate_calls_size_and_time): Likewise.
9258 (estimate_node_size_and_time): Likewise.
9259 (inline_update_callee_summaries): Likewise.
9260 (remap_edge_change_prob): Likewise.
9261 (remap_edge_summaries): Likewise.
9262 (ipa_merge_fn_summary_after_inlining): Likewise.
9263 (ipa_update_overall_fn_summary): Likewise.
9264 (read_ipa_call_summary): Likewise.
9265 (inline_read_section): Likewise.
9266 (write_ipa_call_summary): Likewise.
9267 (ipa_fn_summary_write): Likewise.
9268 (ipa_free_fn_summary): Likewise.
9269 * ipa-hsa.c (process_hsa_functions): Likewise.
9270 (ipa_hsa_write_summary): Likewise.
9271 (ipa_hsa_read_section): Likewise.
9272 * ipa-icf.c (sem_function::merge): Likewise.
9273 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
9274 (do_estimate_edge_time): Likewise.
9275 (estimate_size_after_inlining): Likewise.
9276 (estimate_growth): Likewise.
9277 (growth_likely_positive): Likewise.
9278 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
9279 (inline_call): Likewise.
9280 * ipa-inline.c (caller_growth_limits): Likewise.
9281 (can_inline_edge_p): Likewise.
9282 (can_inline_edge_by_limits_p): Likewise.
9283 (compute_uninlined_call_time): Likewise.
9284 (compute_inlined_call_time): Likewise.
9285 (want_inline_small_function_p): Likewise.
9286 (edge_badness): Likewise.
9287 (update_caller_keys): Likewise.
9288 (update_callee_keys): Likewise.
9289 (recursive_inlining): Likewise.
9290 (inline_small_functions): Likewise.
9291 (inline_to_all_callers_1): Likewise.
9292 (dump_overall_stats): Likewise.
9293 (early_inline_small_functions): Likewise.
9294 (early_inliner): Likewise.
9295 * ipa-inline.h (estimate_edge_growth): Likewise.
9296 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
9297 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9298 * ipa-prop.h (IPA_NODE_REF): Likewise.
9299 (IPA_EDGE_REF): Likewise.
9300 * ipa-pure-const.c (malloc_candidate_p): Likewise.
9301 (propagate_malloc): Likewise.
9302 * ipa-split.c (execute_split_functions): Likewise.
9303 * symbol-summary.h: Rename get to get_create.
9305 (get_create): Likewise.
9306 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
9308 2018-06-08 Martin Liska <mliska@suse.cz>
9310 * symbol-summary.h (release): Move definition out of class
9312 (symtab_removal): Likewise.
9313 (symtab_duplication): Likewise.
9315 2018-06-08 Martin Liska <mliska@suse.cz>
9317 * symbol-summary.h (function_summary): Move constructor
9318 implementation out of class declaration.
9319 (release): Likewise.
9320 (symtab_insertion): Likewise.
9321 (symtab_removal): Likewise.
9322 (symtab_duplication): Likewise.
9325 2018-06-08 Martin Liska <mliska@suse.cz>
9327 * Makefile.in: Remove support for MPX (macros, related functions,
9328 fields in cgraph_node, ...).
9329 * builtin-types.def (BT_BND): Likewise.
9330 (BT_FN_BND_CONST_PTR): Likewise.
9331 (BT_FN_CONST_PTR_BND): Likewise.
9332 (BT_FN_VOID_PTR_BND): Likewise.
9333 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
9334 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
9335 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
9336 (expand_builtin_mempcpy_with_bounds): Likewise.
9337 (expand_builtin_memset_with_bounds): Likewise.
9338 (expand_builtin_memset_args): Likewise.
9339 (std_expand_builtin_va_start): Likewise.
9340 (expand_builtin): Likewise.
9341 (expand_builtin_with_bounds): Likewise.
9342 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
9343 (DEF_LIB_BUILTIN_CHKP): Likewise.
9344 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
9345 (DEF_CHKP_BUILTIN): Likewise.
9346 (BUILT_IN_MEMCPY): Likewise.
9347 (BUILT_IN_MEMMOVE): Likewise.
9348 (BUILT_IN_MEMPCPY): Likewise.
9349 (BUILT_IN_MEMSET): Likewise.
9350 (BUILT_IN_STPCPY): Likewise.
9351 (BUILT_IN_STRCAT): Likewise.
9352 (BUILT_IN_STRCHR): Likewise.
9353 (BUILT_IN_STRCPY): Likewise.
9354 (BUILT_IN_STRLEN): Likewise.
9355 (BUILT_IN_MEMCPY_CHK): Likewise.
9356 (BUILT_IN_MEMMOVE_CHK): Likewise.
9357 (BUILT_IN_MEMPCPY_CHK): Likewise.
9358 (BUILT_IN_MEMSET_CHK): Likewise.
9359 (BUILT_IN_STPCPY_CHK): Likewise.
9360 (BUILT_IN_STRCAT_CHK): Likewise.
9361 (BUILT_IN_STRCPY_CHK): Likewise.
9362 * calls.c (store_bounds): Likewise.
9363 (emit_call_1): Likewise.
9364 (special_function_p): Likewise.
9365 (maybe_warn_nonstring_arg): Likewise.
9366 (initialize_argument_information): Likewise.
9367 (finalize_must_preallocate): Likewise.
9368 (compute_argument_addresses): Likewise.
9369 (expand_call): Likewise.
9370 * cfgexpand.c (expand_call_stmt): Likewise.
9371 (expand_return): Likewise.
9372 (expand_gimple_stmt_1): Likewise.
9373 (pass_expand::execute): Likewise.
9374 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
9375 (cgraph_node::remove): Likewise.
9376 (cgraph_node::dump): Likewise.
9377 (cgraph_node::verify_node): Likewise.
9378 * cgraph.h (chkp_function_instrumented_p): Likewise.
9379 (symtab_node::get_alias_target): Likewise.
9380 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
9381 (cgraph_local_p): Likewise.
9382 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
9383 (cgraph_edge::rebuild_references): Likewise.
9384 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
9385 (walk_polymorphic_call_targets): Likewise.
9386 (cgraph_node::expand_thunk): Likewise.
9387 (symbol_table::output_weakrefs): Likewise.
9388 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
9389 (ix86_handle_option): Likewise.
9390 * config/i386/constraints.md: Likewise.
9391 * config/i386/i386-builtin-types.def (BND): Likewise.
9395 * config/i386/i386-builtin.def (BDESC_END): Likewise.
9396 (BDESC_FIRST): Likewise.
9398 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
9399 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
9400 * config/i386/i386.c (enum reg_class): Likewise.
9401 (ix86_target_string): Likewise.
9402 (ix86_option_override_internal): Likewise.
9403 (ix86_conditional_register_usage): Likewise.
9404 (ix86_valid_target_attribute_inner_p): Likewise.
9405 (ix86_set_indirect_branch_type): Likewise.
9406 (ix86_set_current_function): Likewise.
9407 (ix86_function_arg_regno_p): Likewise.
9408 (init_cumulative_args): Likewise.
9409 (ix86_function_arg_advance): Likewise.
9410 (ix86_function_arg): Likewise.
9411 (ix86_pass_by_reference): Likewise.
9412 (ix86_function_value_regno_p): Likewise.
9413 (ix86_function_value_1): Likewise.
9414 (ix86_function_value_bounds): Likewise.
9415 (ix86_return_in_memory): Likewise.
9416 (ix86_setup_incoming_vararg_bounds): Likewise.
9417 (ix86_va_start): Likewise.
9418 (indirect_thunk_need_prefix): Likewise.
9419 (print_reg): Likewise.
9420 (ix86_print_operand): Likewise.
9421 (ix86_expand_call): Likewise.
9422 (ix86_output_function_return): Likewise.
9423 (reg_encoded_number): Likewise.
9424 (BDESC_VERIFYS): Likewise.
9425 (ix86_init_mpx_builtins): Likewise.
9426 (ix86_init_builtins): Likewise.
9427 (ix86_emit_cmove): Likewise.
9428 (ix86_emit_move_max): Likewise.
9429 (ix86_expand_builtin): Likewise.
9430 (ix86_builtin_mpx_function): Likewise.
9431 (ix86_get_arg_address_for_bt): Likewise.
9432 (ix86_load_bounds): Likewise.
9433 (ix86_store_bounds): Likewise.
9434 (ix86_load_returned_bounds): Likewise.
9435 (ix86_store_returned_bounds): Likewise.
9436 (ix86_class_likely_spilled_p): Likewise.
9437 (ix86_hard_regno_mode_ok): Likewise.
9438 (x86_order_regs_for_local_alloc): Likewise.
9439 (ix86_mitigate_rop): Likewise.
9440 (ix86_bnd_prefixed_insn_p): Likewise.
9441 (ix86_mpx_bound_mode): Likewise.
9442 (ix86_make_bounds_constant): Likewise.
9443 (ix86_initialize_bounds): Likewise.
9444 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
9445 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
9446 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
9447 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
9448 (TARGET_CHKP_BOUND_MODE): Likewise.
9449 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
9450 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
9451 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
9452 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
9453 * config/i386/i386.h (TARGET_MPX): Likewise.
9454 (TARGET_MPX_P): Likewise.
9455 (VALID_BND_REG_MODE): Likewise.
9456 (FIRST_BND_REG): Likewise.
9457 (LAST_BND_REG): Likewise.
9458 (enum reg_class): Likewise.
9459 (BND_REG_P): Likewise.
9460 (BND_REGNO_P): Likewise.
9461 (BNDmode): Likewise.
9462 (ADJUST_INSN_LENGTH): Likewise.
9463 * config/i386/i386.md: Likewise.
9464 * config/i386/i386.opt: Likewise.
9465 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
9466 (defined): Likewise.
9467 (LINK_MPX): Likewise.
9468 (MPX_SPEC): Likewise.
9469 (LIBMPX_SPEC): Likewise.
9470 (LIBMPXWRAPPERS_SPEC): Likewise.
9471 (CHKP_SPEC): Likewise.
9472 * config/i386/predicates.md: Likewise.
9473 * dbxout.c (dbxout_type): Likewise.
9474 * doc/extend.texi: Likewise.
9475 * doc/invoke.texi: Likewise.
9476 * doc/md.texi: Likewise.
9477 * doc/tm.texi: Likewise.
9478 * doc/tm.texi.in: Likewise.
9479 * dwarf2out.c (is_base_type): Likewise.
9480 (gen_formal_types_die): Likewise.
9481 (gen_subprogram_die): Likewise.
9482 (gen_type_die_with_usage): Likewise.
9483 (gen_decl_die): Likewise.
9484 (dwarf2out_late_global_decl): Likewise.
9485 * expr.c (expand_assignment): Likewise.
9486 (emit_storent_insn): Likewise.
9487 (store_expr_with_bounds): Likewise.
9488 (store_expr): Likewise.
9489 (expand_expr_real_1): Likewise.
9490 * expr.h (store_expr_with_bounds): Likewise.
9491 * function.c (use_register_for_decl): Likewise.
9492 (struct bounds_parm_data): Likewise.
9493 (assign_parms_augmented_arg_list): Likewise.
9494 (assign_parm_find_entry_rtl): Likewise.
9495 (assign_parm_is_stack_parm): Likewise.
9496 (assign_parm_load_bounds): Likewise.
9497 (assign_bounds): Likewise.
9498 (assign_parms): Likewise.
9499 (expand_function_start): Likewise.
9500 * gcc.c (CHKP_SPEC): Likewise.
9501 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
9502 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
9503 (wrestrict_dom_walker::check_call): Likewise.
9504 * gimple.c (gimple_build_call_from_tree): Likewise.
9505 * gimple.h (enum gf_mask): Likewise.
9506 (gimple_call_with_bounds_p): Likewise.
9507 (gimple_call_set_with_bounds): Likewise.
9508 * gimplify.c (gimplify_init_constructor): Likewise.
9509 * ipa-cp.c (initialize_node_lattices): Likewise.
9510 (propagate_constants_across_call): Likewise.
9511 (find_more_scalar_values_for_callers_subset): Likewise.
9512 * ipa-hsa.c (process_hsa_functions): Likewise.
9513 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
9514 * ipa-icf.c (sem_function::merge): Likewise.
9515 * ipa-inline.c (early_inliner): Likewise.
9516 * ipa-pure-const.c (warn_function_noreturn): Likewise.
9517 (warn_function_cold): Likewise.
9518 (propagate_pure_const): Likewise.
9519 * ipa-ref.h (enum GTY): Likewise.
9520 * ipa-split.c (find_retbnd): Likewise.
9521 (consider_split): Likewise.
9522 (split_function): Likewise.
9523 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
9524 * ipa.c (walk_polymorphic_call_targets): Likewise.
9525 (symbol_table::remove_unreachable_nodes): Likewise.
9526 (process_references): Likewise.
9527 (cgraph_build_static_cdtor_1): Likewise.
9528 * lto-cgraph.c (lto_output_node): Likewise.
9529 (output_refs): Likewise.
9530 (compute_ltrans_boundary): Likewise.
9531 (input_overwrite_node): Likewise.
9532 (input_node): Likewise.
9533 (input_cgraph_1): Likewise.
9534 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
9535 * passes.c (pass_manager::execute_early_local_passes): Likewise.
9536 (class pass_chkp_instrumentation_passes): Likewise.
9537 (make_pass_chkp_instrumentation_passes): Likewise.
9538 * passes.def: Likewise.
9539 * rtl.h (struct GTY): Likewise.
9540 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
9541 * stor-layout.c (layout_type): Likewise.
9542 * symtab.c: Likewise.
9543 * target.def: Likewise.
9544 * targhooks.c (default_chkp_bound_type): Likewise.
9545 (default_chkp_bound_mode): Likewise.
9546 (default_builtin_chkp_function): Likewise.
9547 (default_chkp_function_value_bounds): Likewise.
9548 (default_chkp_make_bounds_constant): Likewise.
9549 (default_chkp_initialize_bounds): Likewise.
9550 * targhooks.h (default_chkp_bound_type): Likewise.
9551 (default_chkp_bound_mode): Likewise.
9552 (default_builtin_chkp_function): Likewise.
9553 (default_chkp_function_value_bounds): Likewise.
9554 (default_chkp_make_bounds_constant): Likewise.
9555 (default_chkp_initialize_bounds): Likewise.
9556 * toplev.c (compile_file): Likewise.
9557 (process_options): Likewise.
9558 * tree-core.h (DEF_BUILTIN): Likewise.
9559 (DEF_BUILTIN_CHKP): Likewise.
9560 * tree-inline.c (declare_return_variable): Likewise.
9561 (remap_gimple_stmt): Likewise.
9562 (copy_bb): Likewise.
9563 (initialize_inlined_parameters): Likewise.
9564 (expand_call_inline): Likewise.
9565 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
9566 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
9567 (make_pass_ipa_chkp_produce_thunks): Likewise.
9568 (make_pass_chkp): Likewise.
9569 (make_pass_chkp_opt): Likewise.
9570 (make_pass_chkp_instrumentation_passes): Likewise.
9571 * tree-pretty-print.c (dump_generic_node): Likewise.
9572 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
9573 * tree-ssa-dce.c (propagate_necessity): Likewise.
9574 (eliminate_unnecessary_stmts): Likewise.
9575 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
9576 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
9577 * tree-ssa-sccvn.h: Likewise.
9578 * tree-ssa-strlen.c (get_string_length): Likewise.
9579 (valid_builtin_call): Likewise.
9580 (adjust_last_stmt): Likewise.
9581 (handle_builtin_strchr): Likewise.
9582 (handle_builtin_strcpy): Likewise.
9583 (handle_builtin_stxncpy): Likewise.
9584 (handle_builtin_memcpy): Likewise.
9585 (handle_builtin_strcat): Likewise.
9586 (strlen_check_and_optimize_stmt): Likewise.
9587 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
9588 * tree-streamer-in.c: Likewise.
9589 * tree-streamer.c (record_common_node): Likewise.
9590 * tree.c (tree_code_size): Likewise.
9591 (wide_int_to_tree_1): Likewise.
9592 (type_contains_placeholder_1): Likewise.
9593 (build_common_tree_nodes): Likewise.
9594 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
9595 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
9596 (POINTER_BOUNDS_P): Likewise.
9597 (BOUNDED_TYPE_P): Likewise.
9598 (BOUNDED_P): Likewise.
9599 (CALL_WITH_BOUNDS_P): Likewise.
9600 (pointer_bounds_type_node): Likewise.
9601 * value-prof.c (gimple_ic): Likewise.
9602 * var-tracking.c (vt_add_function_parameters): Likewise.
9603 * varasm.c (make_decl_rtl): Likewise.
9604 (assemble_start_function): Likewise.
9605 (output_constant): Likewise.
9606 (maybe_assemble_visibility): Likewise.
9607 * varpool.c (ctor_for_folding): Likewise.
9608 * chkp-builtins.def: Remove.
9609 * ipa-chkp.c: Remove.
9610 * ipa-chkp.h: Remove.
9611 * rtl-chkp.c: Remove.
9612 * rtl-chkp.h: Remove.
9613 * tree-chkp-opt.c: Remove.
9614 * tree-chkp.c: Remove.
9615 * tree-chkp.h: Remove.
9617 2018-06-07 Carl Love <cel@us.ibm.com>
9619 * config/rs6000/vsx.md (vextract_fp_from_shorth,
9620 vextract_fp_from_shortl): Add BE support.
9622 2018-06-07 Paul Koning <ni1d@arrl.net>
9624 * compare-elim.c (try_merge_compare): Don't merge compare if
9625 address contains a side effect.
9626 (try_eliminate_compare): Likewise.
9628 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
9630 * config.gcc: Support "tremont".
9631 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
9632 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
9634 * config/i386/i386.c (m_TREMONT): Define.
9635 (processor_target_table): Add "tremont".
9636 (PTA_TREMONT): Define.
9637 (ix86_lea_outperforms): Add TARGET_TREMONT.
9638 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
9639 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
9640 and M_INTEL_GOLDMONT_PLUS.
9641 (fold_builtin_cpu): Add "tremont".
9642 (ix86_add_stmt_cost): Add TARGET_TREMONT.
9643 (ix86_option_override_internal): Add "tremont".
9644 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
9645 (processor_type): Add PROCESSOR_TREMONT.
9646 * config/i386/x86-tune.def: Add m_TREMONT.
9647 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
9649 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9651 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
9652 symbol defined for msp430i* devices to be lower case.
9654 2018-06-07 Richard Biener <rguenther@suse.de>
9656 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
9657 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
9658 Properly wrap signed arithmetic if overflow wraps.
9660 2018-06-07 Jakub Jelinek <jakub@redhat.com>
9662 PR tree-optimization/69615
9663 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
9664 of a cast from a same precision integral SSA_NAME in a bb dominated
9665 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
9666 cast to utype if rhs2 has already a compatible type.
9668 2018-06-07 Richard Biener <rguenther@suse.de>
9670 PR tree-optimization/85935
9671 * graphite-scop-detection.c (find_params_in_bb): Analyze
9672 condition operands with respect to the correct loop. Assert
9673 the analysis doesn't fail.
9675 2018-06-04 Carl Love <cel@us.ibm.com>
9677 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
9678 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
9679 as it is slightly cheaper.
9680 (first_match_or_eos_index_<mode>):
9681 Calculate index using natural element order.
9682 (first_match_index_<mode>):
9683 Calculate index using natural element order.
9684 (first_match_or_eos_index_<mode>):
9685 Calculate index using natural order.
9686 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
9687 for BE and LE modes.
9688 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
9689 P9V_BUILTIN_VCLZLSBB_V16QI.
9690 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
9693 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
9695 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
9696 indentation and line wrap for many prototypes. Add missing
9697 @smallexample directives around block of prototypes for vec_xl and
9700 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
9702 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
9703 track if we pass or return IEEE 128-bit floating point.
9704 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
9705 C++ mangling that is compatible with GCC 8.1.
9706 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
9707 (init_cumulative_args): Note if we pass or return IEEE 128-bit
9708 floating point types.
9709 (rs6000_function_arg_advance_1): Likewise.
9710 (rs6000_mangle_type): Optionally generate mangled names that match
9711 what GCC 8.1 generated for IEEE 128-bit floating point types.
9712 (rs6000_globalize_decl_name): If we have an external function that
9713 passes or returns IEEE 128-bit types, generate a weak reference
9714 from the mangled name used in GCC 8.1 to the current mangled
9716 (rs6000_init_builtins): Make __ibm128 use the long double type if
9717 long double is IBM extended double. Make __float128 use the long
9718 double type if long double is IEEE 128-bit.
9721 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
9722 macro for __ibm128 built-in functions.
9723 (PACK_IF): Add __ibm128 pack/unpack functions.
9724 (UNPACK_IF): Likewise.
9725 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
9726 enable long double built-in functions if long double is IEEE
9727 128-bit floating point.
9728 (rs6000_invalid_builtin): Update long double built-in function
9730 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
9731 functions, adjust the built-in function to use the long double
9732 built-in function if __ibm128 and long double are the same type.
9733 * doc/extend.texi (PowerPC builtins): Update documention for
9734 __builtin_{,un}pack_longdouble. Add documentation for
9735 __builtin_{,un}pack_ibm128.
9737 2018-06-06 Jim Wilson <jimw@sifive.com>
9739 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
9740 (struct machine_function): New field interrupt_mode.
9741 (riscv_handle_type_attribute): New function. Add forward declaration.
9742 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
9743 (riscv_expand_epilogue): Check interrupt_mode field.
9744 (riscv_set_current_function): Check interrupt attribute args and
9745 set interrupt_mode field.
9746 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
9747 (riscv_sret, riscv_uret): New.
9748 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
9749 new arguments to interrupt attribute.
9751 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
9754 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
9755 Don't handle -mcpu=power8 if -mpower9-vector is also used.
9757 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
9759 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
9760 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
9761 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9762 several redundant entries.
9764 2018-06-06 David Malcolm <dmalcolm@redhat.com>
9766 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
9767 type from "rtx" to "rtx_insn *".
9768 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
9769 for local "call_insn", removing cast.
9770 (ix86_expand_call): Likewise, introducing a "call_insn" local.
9772 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
9774 PR tree-optimization/86066
9775 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
9776 for BIT_INSERT_EXPR stores.
9778 2018-06-06 Richard Biener <rguenther@suse.de>
9780 PR tree-optimization/86062
9781 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
9782 component refs ontop
9783 of to be offsetted base.
9785 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9787 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
9788 to be static and remove check on interrupt attribute name.
9790 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
9792 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
9793 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
9795 2018-06-05 Steve Ellcey <sellcey@cavium.com>
9798 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
9800 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
9801 Remove second argument, change how error is called.
9802 (aarch64_layout_arg): Remove second argument from
9803 aarch64_err_no_fpadvsimd call.
9804 (aarch64_init_cumulative_args): Ditto.
9805 (aarch64_gimplify_va_arg_expr): Ditto.
9806 * config/aarch64/aarch64.md (mov<mode>): Ditto.
9808 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
9810 * config/i386/i386.md (simple_return_indirect_internal): New expander.
9811 (*simple_return_indirect_internal<mode>): Rename from
9812 simple_return_indirect_internal. Use W mode iterator.
9813 (rstorssp): New expander.
9814 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
9815 (clrssbsy): New expander.
9816 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
9818 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
9820 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
9822 (cmse_check_pointed_object): Likewise.
9824 2018-06-05 Martin Liska <mliska@suse.cz>
9826 PR gcov-profile/47618
9827 * doc/invoke.texi: Document how -fprofile-dir format
9830 2018-06-05 Richard Biener <rguenther@suse.de>
9832 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
9833 removal pretend DOM info isn't available so we do not update
9834 it and only remove edges, not dominated blocks. Actually free
9835 DOM info in case we removed something. Remove unreachable blocks.
9836 (mfb_keep_latches): Work with either DOM info or marked backedges.
9837 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
9838 first. Mark backedges if DOM info isn't available.
9839 (Re-)compute DOM info after cleanup_control_flow_pre.
9841 2018-06-05 Richard Biener <rguenther@suse.de>
9843 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
9844 (locus_discrim_hasher::hash): Adjust.
9845 (locus_discrim_hasher::equal): Likewise.
9846 (next_discriminator_for_locus): Work on line directly.
9847 (same_line_p): Pass in expanded locus1 as well.
9848 (assign_discriminators): Avoid redundant location expansions.
9850 2018-06-05 Richard Biener <rguenther@suse.de>
9852 PR tree-optimization/86046
9853 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
9854 if required after clearing TREE_ADDRESSABLE.
9856 2018-06-05 Richard Biener <rguenther@suse.de>
9858 PR tree-optimization/86047
9859 * tree-ssa-loop.c (for_each_index): Glob handling of all
9860 decls and constants and really handle all of them.
9862 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9865 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
9866 qualifier_void_pointer and qualifier_const_void_pointer.
9867 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
9868 (arm_init_builtins): Handle the above.
9869 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
9870 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
9871 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
9874 2018-06-05 Martin Liska <mliska@suse.cz>
9876 * auto-profile.c (read_autofdo_file): Do not use
9877 gcov_ctr_summary struct.
9878 (afdo_callsite_hot_enough_for_early_inline): Likewise.
9879 * coverage.c (struct counts_entry): Likewise.
9880 (read_counts_file): Read just single summary entry.
9881 (get_coverage_counts): Use gcov_summary struct.
9882 * coverage.h (get_coverage_counts): Likewise.
9883 * gcov-dump.c (dump_working_sets): Likewise.
9884 (tag_summary): Dump just single summary.
9885 * gcov-io.c (gcov_write_summary): Write just histogram
9887 (gcov_read_summary): Read just single summary.
9888 (compute_working_sets): Use gcov_summary struct.
9889 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
9890 of GCOV_COUNTERS_SUMMABLE.
9891 (GCOV_COUNTERS_SUMMABLE): Remove.
9892 (GCOV_FIRST_VALUE_COUNTER): Replace with
9893 GCOV_COUNTER_V_INTERVAL.
9894 (struct gcov_ctr_summary): Remove.
9895 (struct gcov_summary): Directly use fields of former
9897 (compute_working_sets): Use gcov_summary struct.
9898 * gcov.c (read_count_file): Do not use ctrs fields.
9899 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
9901 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
9903 * profile.c: Likewise.
9904 * profile.h: Likewise.
9906 2018-06-05 Martin Liska <mliska@suse.cz>
9908 PR gcov-profile/84846
9909 * gcov.c (output_lines): Print working directory only
9910 in intermediate format.
9912 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
9914 * config/s390/s390-builtin-types.def: Add void function type.
9915 * config/s390/s390-builtins.def: Use the function type for the
9918 2018-06-04 Jim Wilson <jimw@sifive.com>
9920 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
9922 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
9923 and maybe_eh_return. Change regno to unsigned int. Use new args to
9924 handle EH_RETURN_DATA_REGNO registers properly.
9925 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
9926 (riscv_expand_epilogue): Update comment. Change argument name and
9927 type. Update code to use new name and type. Pass new args to
9928 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
9930 * config/riscv/riscv.md (NORMAL_RETURN): New.
9931 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
9932 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
9933 (eh_return): Call gen_eh_return_internal and emit barrier.
9934 (eh_return_internal): Call riscv_expand_epilogue.
9936 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
9938 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
9939 bit_insertion field and declare can_be_merged_into method.
9940 (merged_store_group::can_be_merged_into): New method.
9941 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
9942 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
9943 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
9945 2018-06-04 Richard Biener <rguenther@suse.de>
9947 PR tree-optimization/85955
9948 * builtins.c (fold_builtin_sincos): Convert pointers to
9949 destination to appropriate type before dereferencing.
9951 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
9953 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
9955 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
9957 * expr.c (expand_expr_real_1): Force the operand into memory if
9958 its TYPE_MODE is BLKmode and if there is no integer mode for
9959 the number of bits being extracted.
9961 2018-06-04 Jakub Jelinek <jakub@redhat.com>
9965 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
9966 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
9968 2018-06-04 Richard Biener <rguenther@suse.de>
9970 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
9971 (cleanup_tree_cfg_noloop): ... single caller. Do
9972 start_recording_case_labels later.
9974 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
9976 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
9977 to _IMMINTRIN_H_INCLUDED.
9978 * config/i386/pconfigintrin.h: Ditto.
9979 * config/i386/waitpkgintrin.h: Ditto.
9980 * config/i386/immintrin.h: Add includes for sgxintrin.h,
9981 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
9982 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
9983 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
9984 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
9985 waitpkgintrin.h and cldemoteintrin.h.
9987 2018-06-04 Richard Biener <rguenther@suse.de>
9989 PR tree-optimization/86038
9990 * tracer.c (find_best_successor): Check probability for
9991 being initialized, bail out if not.
9993 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
9996 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
9997 of bits to ignore when comparing architectures.
9999 2018-06-04 Jakub Jelinek <jakub@redhat.com>
10001 PR tree-optimization/69615
10002 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
10003 maximum or minimum of the type, try to merge it also as if
10004 range1 is + [-, x - 1] or + [x + 1, -].
10007 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
10009 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
10011 PR tree-optimization/86034
10012 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
10013 the unsigned bitfield type in a bit insertion sequence if it does not
10014 have a larger precision than the bitfield size.
10015 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
10017 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
10019 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
10021 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
10023 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
10024 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
10025 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
10026 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
10028 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
10030 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
10031 Disable -fdelete-null-pointer-checks for ELF toolchain.
10033 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
10034 Kito Cheng <kito.cheng@gmail.com>
10036 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
10037 (nds32le-*-*, nds32be-*-*): Integrate checking process.
10038 (nds32*-*-*): Add glibc and uclibc conditions.
10039 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
10040 (TARGET_EXCEPT_UNWIND_INFO): Define.
10041 * config/nds32/elf.h: New file.
10042 * config/nds32/linux.h: New file.
10043 * config/nds32/nds32-elf.opt: New file.
10044 * config/nds32/nds32-linux.opt: New file.
10045 * config/nds32/nds32-fp-as-gp.c
10046 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
10047 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
10049 (nds32_asm_file_end): Ditto.
10050 (nds32_print_operand): Ditto.
10051 (nds32_insert_attributes): Ditto.
10052 (nds32_init_libfuncs): New function.
10053 (TARGET_HAVE_TLS): Define.
10054 (TARGET_INIT_LIBFUNCS): Define.
10055 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
10057 (TARGET_ELF): Apply different mcmodel setting.
10058 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
10059 been migrated into elf.h and linux.h files.
10060 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
10061 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
10062 (mcmodel): The content has been migrated into nds32-elf.opt and
10063 nds32-linux.opt files.
10064 * config/nds32/t-elf: New file.
10065 * config/nds32/t-linux: New file.
10067 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
10068 Shiva Chen <shiva0217@gmail.com>
10070 * config/nds32/constants.md (unspec_volatile_element): Add
10071 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
10072 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
10074 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
10075 (make_pass_nds32_fp_as_gp): Declare.
10076 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
10078 (nds32_asm_function_end_prologue): Remove unused asm output.
10079 (nds32_asm_function_begin_epilogue): Remove unused asm output.
10080 (nds32_asm_file_start): Output necessary fp_as_gp information.
10081 (nds32_option_override): Adjust register usage.
10082 (nds32_expand_prologue): Consider fp_as_gp situation.
10083 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
10084 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
10087 (simple_return): Ditto.
10088 (omit_fp_begin): Output special directive for fp_as_gp.
10089 (omit_fp_end): Output special directive for fp_as_gp.
10090 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
10091 mforbid-fp-as-gp): New options.
10093 2018-06-01 Mark Wielaard <mark@klomp.org>
10095 * dwarf2out.c (dwarf2out_finish): Remove generation of
10096 DW_AT_loclists_base.
10098 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
10100 * gimple-ssa-store-merging.c: Include gimple-fold.h.
10101 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
10102 (struct merged_store_group): Add bit_insertion field.
10103 (dump_char_array): Use standard hexadecimal format.
10104 (merged_store_group::merged_store_group): Set bit_insertion to false.
10105 (merged_store_group::apply_stores): Use optimal buffer size. Deal
10106 with BIT_INSERT_EXPR stores. Move up code updating the mask and
10107 also print the mask in the dump file.
10108 (pass_store_merging::gate): Minor tweak.
10109 (imm_store_chain_info::coalesce_immediate): Fix wrong association
10110 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
10111 stores with INTEGER_CST stores.
10112 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
10113 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
10114 and use it throughout. Generate bit insertion sequences if need be.
10115 (pass_store_merging::process_store): Remove redundant condition.
10116 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
10118 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
10120 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
10121 the 128-bit floating point types. Fix function comment.
10123 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10125 * config/aarch64/aarch64-simd.md
10126 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
10128 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
10131 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
10133 PR tree-optimization/85989
10134 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
10136 (backprop::intersect_uses): Check it when deciding whether this
10137 is a backedge reference.
10138 (backprop::process_block): Add each phi to m_visited_phis
10139 after visiting it, then clear it at the end.
10141 2018-06-01 Richard Biener <rguenther@suse.de>
10143 * tree-vectorizer.h (vect_dr_stmt): New function.
10144 (vect_get_load_cost): Adjust.
10145 (vect_get_store_cost): Likewise.
10146 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
10147 Use vect_dr_stmt instead of DR_SMTT.
10148 (vect_record_base_alignments): Likewise.
10149 (vect_calculate_target_alignment): Likewise.
10150 (vect_compute_data_ref_alignment): Likewise and make static.
10151 (vect_update_misalignment_for_peel): Likewise.
10152 (vect_verify_datarefs_alignment): Likewise.
10153 (vector_alignment_reachable_p): Likewise.
10154 (vect_get_data_access_cost): Likewise. Pass down
10155 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
10156 (vect_get_peeling_costs_all_drs): Likewise.
10157 (vect_peeling_hash_get_lowest_cost): Likewise.
10158 (vect_enhance_data_refs_alignment): Likewise.
10159 (vect_find_same_alignment_drs): Likewise.
10160 (vect_analyze_data_refs_alignment): Likewise.
10161 (vect_analyze_group_access_1): Likewise.
10162 (vect_analyze_group_access): Likewise.
10163 (vect_analyze_data_ref_access): Likewise.
10164 (vect_analyze_data_ref_accesses): Likewise.
10165 (vect_vfa_segment_size): Likewise.
10166 (vect_small_gap_p): Likewise.
10167 (vectorizable_with_step_bound_p): Likewise.
10168 (vect_prune_runtime_alias_test_list): Likewise.
10169 (vect_analyze_data_refs): Likewise.
10170 (vect_supportable_dr_alignment): Likewise.
10171 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10172 (vect_gen_prolog_loop_niters): Likewise.
10173 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10174 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
10176 (vect_recog_mask_conversion_pattern): Likewise.
10177 (vect_try_gather_scatter_pattern): Likewise.
10178 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
10179 to vect_get_store_cost.
10180 (vect_get_store_cost): Get stmt_info instead of DR.
10181 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
10182 (vect_get_load_cost): Get stmt_info instead of DR.
10184 2018-06-01 Richard Biener <rguenther@suse.de>
10186 PR middle-end/86017
10187 * gimple-fold.c (var_decl_component_p): Also allow offsetted
10188 vars wrapped in MEM_REFs.
10190 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
10192 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
10193 Fix subreg tests so that we only return a choice between
10194 GENERAL_REGS and FP_REGS if the original classes included both.
10196 2018-06-01 Richard Biener <rguenther@suse.de>
10199 * tree-ssa-structalias.c (get_function_part_constraint):
10200 Handle NULL fi->decl.
10201 (find_func_aliases_for_call): Properly handle indirect
10202 fi from direct call.
10203 (find_func_clobbers): Likewise.
10204 (ipa_pta_execute): Likewise.
10205 (create_variable_info_for): For functions that are ifunc_resolver
10206 resolve to a varinfo that contains the result of the resolver call.
10207 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
10210 2018-05-31 Michael Collison <michael.collison@arm.com>
10212 * config/aarch64/aarch64.md:
10213 (*fix_to_zero_extenddfdi2): New pattern.
10214 * gcc.target/aarch64/fix_extend1.c: New testcase.
10216 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
10218 PR middle-end/78809
10219 PR middle-end/83026
10220 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
10221 and BUILT_IN_STRNCMP_EQ.
10222 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
10223 BUILT_IN_STRNCMP_EQ.
10224 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
10225 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
10226 (gimple_fold_builtin): Likewise.
10227 * tree-ssa-strlen.c (compute_string_length): New function.
10228 (determine_min_obsize): New function.
10229 (handle_builtin_string_cmp): New function to handle calls to
10230 string compare functions.
10231 (strlen_optimize_stmt): Add handling to builtin string compare
10233 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
10234 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
10235 * tree.c (build_common_builtin_nodes): Add new defines of
10236 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
10238 2018-05-31 Jakub Jelinek <jakub@redhat.com>
10241 * bb-reorder.c (pass_partition_blocks::gate): Return false for
10242 functions with naked attribute.
10244 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
10246 * config/i386/sse.md (avx_vec_concat<mode>):
10247 Substitute concat_tg_mode mode attribute with xtg_mode.
10248 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
10249 (concat_tg_mode): Remove mode attribute.
10251 2018-05-31 Martin Sebor <msebor@redhat.com>
10254 * calls.c (alloc_max_size): Correct a logic error/typo.
10255 Treat excessive arguments as infinite. Warn for invalid arguments.
10256 * doc/invoke.texi (-Walloc-size-larger-than): Update.
10258 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
10261 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
10262 and movx for Haswell.
10264 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
10265 Cesar Philippidis <cesar@codesourcery.com>
10267 PR middle-end/85879
10268 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
10269 when emitting error on private/firstprivate reductions.
10270 * omp-low.c (lower_omp_target): Avoid reference-type processing
10271 on pointers for firstprivate clause.
10273 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
10275 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
10278 * config/aarch64/aarch64-simd.md
10279 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
10280 (aarch64_ld1_x3_<mode>): Likewise
10281 (aarch64_st1x2<VALLDIF:mode>): Likewise
10282 (aarch64_st1_x2_<mode>): Likewise
10283 (aarch64_st1x3<VALLDIF:mode>): Likewise
10284 (aarch64_st1_x3_<mode>): Likewise
10285 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
10286 (vld1_s8_x3): Likewise.
10287 (vld1_u16_x3): Likewise.
10288 (vld1_s16_x3): Likewise.
10289 (vld1_u32_x3): Likewise.
10290 (vld1_s32_x3): Likewise.
10291 (vld1_u64_x3): Likewise.
10292 (vld1_s64_x3): Likewise.
10293 (vld1_f16_x3): Likewise.
10294 (vld1_f32_x3): Likewise.
10295 (vld1_f64_x3): Likewise.
10296 (vld1_p8_x3): Likewise.
10297 (vld1_p16_x3): Likewise.
10298 (vld1_p64_x3): Likewise.
10299 (vld1q_u8_x3): Likewise.
10300 (vld1q_s8_x3): Likewise.
10301 (vld1q_u16_x3): Likewise.
10302 (vld1q_s16_x3): Likewise.
10303 (vld1q_u32_x3): Likewise.
10304 (vld1q_s32_x3): Likewise.
10305 (vld1q_u64_x3): Likewise.
10306 (vld1q_s64_x3): Likewise.
10307 (vld1q_f16_x3): Likewise.
10308 (vld1q_f32_x3): Likewise.
10309 (vld1q_f64_x3): Likewise.
10310 (vld1q_p8_x3): Likewise.
10311 (vld1q_p16_x3): Likewise.
10312 (vld1q_p64_x3): Likewise.
10313 (vst1_s64_x2): Likewise.
10314 (vst1_u64_x2): Likewise.
10315 (vst1_f64_x2): Likewise.
10316 (vst1_s8_x2): Likewise.
10317 (vst1_p8_x2): Likewise.
10318 (vst1_s16_x2): Likewise.
10319 (vst1_p16_x2): Likewise.
10320 (vst1_s32_x2): Likewise.
10321 (vst1_u8_x2): Likewise.
10322 (vst1_u16_x2): Likewise.
10323 (vst1_u32_x2): Likewise.
10324 (vst1_f16_x2): Likewise.
10325 (vst1_f32_x2): Likewise.
10326 (vst1_p64_x2): Likewise.
10327 (vst1q_s8_x2): Likewise.
10328 (vst1q_p8_x2): Likewise.
10329 (vst1q_s16_x2): Likewise.
10330 (vst1q_p16_x2): Likewise.
10331 (vst1q_s32_x2): Likewise.
10332 (vst1q_s64_x2): Likewise.
10333 (vst1q_u8_x2): Likewise.
10334 (vst1q_u16_x2): Likewise.
10335 (vst1q_u32_x2): Likewise.
10336 (vst1q_u64_x2): Likewise.
10337 (vst1q_f16_x2): Likewise.
10338 (vst1q_f32_x2): Likewise.
10339 (vst1q_f64_x2): Likewise.
10340 (vst1q_p64_x2): Likewise.
10341 (vst1_s64_x3): Likewise.
10342 (vst1_u64_x3): Likewise.
10343 (vst1_f64_x3): Likewise.
10344 (vst1_s8_x3): Likewise.
10345 (vst1_p8_x3): Likewise.
10346 (vst1_s16_x3): Likewise.
10347 (vst1_p16_x3): Likewise.
10348 (vst1_s32_x3): Likewise.
10349 (vst1_u8_x3): Likewise.
10350 (vst1_u16_x3): Likewise.
10351 (vst1_u32_x3): Likewise.
10352 (vst1_f16_x3): Likewise.
10353 (vst1_f32_x3): Likewise.
10354 (vst1_p64_x3): Likewise.
10355 (vst1q_s8_x3): Likewise.
10356 (vst1q_p8_x3): Likewise.
10357 (vst1q_s16_x3): Likewise.
10358 (vst1q_p16_x3): Likewise.
10359 (vst1q_s32_x3): Likewise.
10360 (vst1q_s64_x3): Likewise.
10361 (vst1q_u8_x3): Likewise.
10362 (vst1q_u16_x3): Likewise.
10363 (vst1q_u32_x3): Likewise.
10364 (vst1q_u64_x3): Likewise.
10365 (vst1q_f16_x3): Likewise.
10366 (vst1q_f32_x3): Likewise.
10367 (vst1q_f64_x3): Likewise.
10368 (vst1q_p64_x3): Likewise.
10370 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10372 * config/msp430/msp430.c (msp430_output_labelref): Prepend
10373 user_label_prefix to name.
10375 * tree-core.h: Update comment about the format of NAME string
10376 passed to handler in attribute_spec.
10378 * config/msp430/msp430.md: Remove erroneous subreg expression from
10379 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
10380 zero_extend{q,h}isi2.
10382 2018-05-30 Borislav Petkov <bp@suse.de>
10384 * doc/extend.texi: Document some architecture specific
10385 constraints and sort entries.
10387 2018-05-30 Martin Sebor <msebor@redhat.com>
10389 PR middle-end/85369
10390 * builtins.c (expand_builtin_stpcpy_1): New function.
10391 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
10392 only if the former succeeds.
10394 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
10396 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
10399 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10401 * doc/invoke.texi (-flinker-output): Document
10403 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10405 * passes.c (ipa_write_summaries): Only modify statements if body
10407 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
10408 incrementally linking.
10409 (ipa_passes): Likewise.
10410 * lto-cgraph.c (lto_output_node): When incrementally linking do not
10411 pass down resolution info.
10412 * common.opt (flag_incremental_link): Update info.
10413 * gcc.c (plugin specs): Turn flinker-output=* to
10414 -plugin-opt=-linker-output-known
10415 * toplev.c (compile_file): Also cut compilation when doing incremental
10417 * flag-types. (enum lto_partition_model): Add
10418 LTO_LINKER_OUTPUT_NOLTOREL.
10419 (invoke.texi): Add -flinker-output docs.
10420 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
10421 link same way as WPA; do not stream in dead initializers.
10423 * dwarf2out.c (dwarf2out_die_ref_for_decl,
10424 darf2out_register_external_decl): Support incremental link.
10426 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10428 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
10430 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10432 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
10433 it down to simple_object_copy_lto_debug_sections.
10434 (run_gcc): Determine incremental LTO link time and configure
10435 lto1 into non-wpa mode, disable renaming of debug sections.
10437 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
10439 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
10440 descriptions of various incorrectly documented functions.
10442 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
10445 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
10446 address check not strict.
10448 2018-05-30 Richard Biener <rguenther@suse.de>
10450 PR tree-optimization/85964
10451 * tracer.c (better_p): Drop initialized count check, we only
10452 call the function with initialized counts now.
10453 (find_best_successor): Do find a best edge if one
10454 has uninitialized count.
10455 (find_best_predecessor): Likewise. Do BB frequency check only
10456 if count is initialized.
10458 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
10460 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
10461 (aarch64_ldrstr_offset_compare): New.
10462 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
10463 load/store orderings.
10464 (aarch64_gen_adjusted_ldpstp): Likewise.
10466 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
10468 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
10469 Check for subset of GENERAL_REGS and FP_REGS.
10470 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
10473 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
10475 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
10476 and wi::to_poly_offset. Add the current offset and then check
10477 whether the sum fits, rather than using an unchecked addition of
10478 a checked term. Check for a shwi rather than a uhwi.
10479 * expr.c (get_bit_range): Use tree_to_poly_uint64.
10480 (store_constructor): Use poly_int_tree_p.
10481 (expand_expr_real_1): Likewise.
10482 * function.c (assign_temp): Likewise.
10483 * fold-const.c (const_binop): Use poly_int_tree_p and
10484 wi::to_poly_offset.
10485 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
10487 * ipa-icf-gimple.c (func_checker::compare_operand): Use
10488 to_poly_offset for MEM offsets.
10489 * ipa-icf.c (sem_variable::equals): Likewise.
10490 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
10491 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
10492 wi::to_poly_offset for BIT_FIELD_REF offsets.
10493 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
10494 wi::to_poly_offset.
10495 * var-tracking.c (emit_note_insn_var_location): Use
10496 tree_to_poly_uint64.
10498 2018-05-29 Jim Wilson <jimw@sifive.com>
10500 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
10502 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
10505 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
10506 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
10508 (sse4_1_round<mode>2): Use nonimmediate_operand
10509 for operand 1 predicate.
10511 2018-05-29 Martin Sebor <msebor@redhat.com>
10512 Richard Biener <rguenther@suse.de>
10515 * calls.c (get_size_range): Call determine_value_range instead
10516 of get_value_range..
10517 * tree-vrp.h (determine_value_range): Declared new function.
10518 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
10520 2018-05-29 Richard Biener <rguenther@suse.de>
10522 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
10523 sure to use non-pattern stmts for get_earlier_stmt arguments.
10524 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
10525 called on pattern stmts.
10526 (get_later_stmt): Likewise.
10528 2018-05-29 Martin Liska <mliska@suse.cz>
10530 PR gcov-profile/85759
10531 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
10534 2018-05-29 Jakub Jelinek <jakub@redhat.com>
10536 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
10538 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
10542 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
10543 VEC_PACK_FLOAT_EXPR): New tree codes.
10544 * tree-pretty-print.c (op_code_prio): Handle
10545 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
10546 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
10547 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
10548 * tree-inline.c (estimate_operator_cost): Likewise.
10549 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
10550 * fold-const.c (const_binop): Likewise.
10551 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
10552 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
10553 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
10554 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
10555 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
10556 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
10557 * expr.c (expand_expr_real_2): Likewise.
10558 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
10559 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
10560 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
10562 * optabs.c (expand_widen_pattern_expr): For
10563 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
10564 sign from result type rather than operand's type.
10565 (expand_binop_directly): For vec_packu_float_optab and
10566 vec_packs_float_optab allow result type to be different from operand's
10568 * optabs-tree.c (optab_for_tree_code): Handle
10569 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
10570 VEC_PACK_FLOAT_EXPR. Formatting fixes.
10571 * tree-vect-generic.c (expand_vector_operations_1): Handle
10572 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
10573 VEC_PACK_FLOAT_EXPR.
10574 * tree-vect-stmts.c (supportable_widening_operation): Handle
10576 (supportable_narrowing_operation): Handle FLOAT_EXPR.
10577 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
10578 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
10579 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
10580 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
10582 (vec_pack<floatprefix>_float_<mode>): New expander.
10583 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
10585 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
10586 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
10587 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
10588 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
10589 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
10591 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
10592 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
10593 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
10594 VEC_PACK_FLOAT_EXPR): Document.
10596 2018-05-29 Richard Biener <rguenther@suse.de>
10598 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
10600 (stmt_vec_info_vec): Make pointer.
10601 (init_stmt_vec_info_vec): Remove.
10602 (free_stmt_vec_info_vec): Likewise.
10603 (set_stmt_vec_info_vec): New function.
10604 (free_stmt_vec_infos): Likewise.
10605 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
10606 (set_vinfo_for_stmt): Likewise.
10607 (get_earlier_stmt): Likewise.
10608 (get_later_stmt): Likewise.
10609 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
10610 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
10611 (vec_info::~vec_info): Free stmt_vec_infos.
10612 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
10613 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
10614 (pass_slp_vectorize::execute): Likewise.
10615 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
10616 (free_stmt_vec_info_vec): Likewise.
10617 (set_stmt_vec_info_vec): New function.
10618 (free_stmt_vec_infos): Likewise.
10619 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
10620 the global stmt_vec_info_vec.
10621 * tree-parloops.c (gather_scalar_reductions): Use
10622 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
10625 2018-05-29 Richard Biener <rguenther@suse.de>
10627 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
10629 2018-05-29 Martin Liska <mliska@suse.cz>
10630 David Malcolm <dmalcolm@redhat.com>
10632 * vec.c (test_reverse): New.
10633 (vec_c_tests): Add new test.
10634 * vec.h (vl_ptr>::reverse): New function.
10636 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
10638 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
10640 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
10643 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
10645 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
10647 2018-05-28 Richard Biener <rguenther@suse.de>
10649 PR tree-optimization/85933
10650 * tree-vect-data-refs.c (vect_record_base_alignments): Only
10651 look at stmts marked as vectorizable.
10653 2018-05-28 Richard Biener <rguenther@suse.de>
10655 PR tree-optimization/85934
10656 * tree-vect-generic.c (expand_vector_operations_1): Hoist
10657 vector boolean check before scalar optimization.
10659 2018-05-28 Jakub Jelinek <jakub@redhat.com>
10661 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
10664 2018-05-28 Mark Wielaard <mark@klomp.org>
10666 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
10667 if it is an expression containing a minus sign.
10669 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
10671 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
10673 2018-05-27 Paul Koning <ni1d@arrl.net>
10675 * config/pdp11/pdp11.md (truncsihi2): Remove.
10677 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
10678 Chung-Ju Wu <jasonwucj@gmail.com>
10680 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
10682 (unaligned_store_dw): Ditto.
10683 * config/nds32/nds32-memory-manipulation.c
10684 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
10685 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
10686 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
10687 (emit_setmem_word_loop): Rename to ...
10688 (emit_setmem_doubleword_loop): ... this.
10689 (nds32_gen_dup_4_byte_to_word_value): New function.
10690 (nds32_gen_dup_8_byte_to_double_word_value): New function.
10691 (nds32_expand_setmem_loop): Refine implementation.
10692 (nds32_expand_setmem_loop_v3m): Ditto.
10693 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
10696 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
10698 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
10700 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
10702 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
10703 (nds32_init_machine_status): Initialize machine->attr_naked_p and
10704 machine->attr_no_prologue_p.
10705 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
10706 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
10707 (nds32_expand_epilogue): Consider attr_naked_p.
10708 (nds32_expand_epilogue_v3pop): Likewise.
10709 (nds32_can_use_return_insn): Likewise.
10710 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
10711 attr_no_prologue_p fields.
10712 * config/nds32/nds32.opt (mret-in-naked-func): New option.
10714 2018-05-27 Jakub Jelinek <jakub@redhat.com>
10717 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
10719 * config/i386/sse.md
10720 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
10722 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
10724 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
10726 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
10728 (*<floatsuffix>floatv2div2sf2): Rename to ...
10729 (*float<floatunssuffix>v2div2sf2): ... this.
10730 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
10731 (float<floatunssuffix>v2div2sf2_mask): ... this.
10732 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
10733 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
10734 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
10736 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
10738 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
10740 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
10742 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
10744 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
10746 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
10747 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
10748 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
10749 gen_ufix_truncv8dfv8si2.
10750 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
10751 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
10752 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
10753 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
10754 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
10755 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
10756 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
10757 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
10759 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
10763 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
10765 2018-05-25 Jim Wilson <jimw@sifive.com>
10767 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
10768 * config/riscv/riscv.c (struct machine_function): Add
10769 interrupt_handler_p and attribute_checked_p fields.
10770 (riscv_attribute_table): Add interrupt.
10771 (riscv_interrupt_type_p): New.
10772 (riscv_save_reg_p): Save extra regs for interrupt handler.
10773 (riscv_use_save_libcall): Return false for interrupt handler.
10774 (riscv_first_stack_step): Add forward declaration.
10775 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
10776 for interrupt handler with large frame. Use it for saved reg list.
10777 (riscv_expand_prologue): Move flag_stack_usage_info support to
10778 eliminate duplication.
10779 (riscv_expand_epilogue): Generate mret for interrupt handler.
10780 (riscv_epilogue_uses): New.
10781 (riscv_can_use_return_insn): Return false for interrupt handler.
10782 (riscv_function_ok_for_sibcall): Likewise.
10783 (riscv_set_current_function): Add interrupt handler support.
10784 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
10785 * config/riscv/riscv.md (UNSPECV_MRET): New.
10787 (riscv_frflags, riscv_fsflags): Use tab after opcode.
10789 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
10791 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
10793 PR tree-optimization/85712
10794 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
10795 this candidate has already been replaced in-situ by a copy.
10797 2018-05-25 Jason Merrill <jason@redhat.com>
10799 PR c++/80485 - inline function non-zero address.
10800 * symtab.c (nonzero_address): Check DECL_COMDAT.
10802 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
10805 * config/alpha/alpha.md (ashlsi3): New insn pattern.
10806 (*ashlsi_se): Rename from *ashldi_se. Define as sign
10807 extension of SImode operation. Use const123_operand predicate.
10808 (*saddsi_1): Remove.
10809 (*saddl_se_1): Ditto.
10810 (*ssubsi_1): Ditto.
10811 (*ssubl_se_1): Ditto.
10812 * config/alpha/predicates.md (const123_operand): New predicate.
10813 * config/alpha/constraints.md (P): Use IN_RANGE.
10815 2018-05-25 Richard Biener <rguenther@suse.de>
10817 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
10819 (ref_maybe_used_by_stmt_p): Likewise.
10820 (stmt_may_clobber_ref_p): Likewise.
10821 (stmt_may_clobber_ref_p_1): Likewise.
10822 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
10824 (ref_maybe_used_by_stmt_p): Likewise.
10825 (stmt_may_clobber_ref_p): Likewise.
10826 (stmt_may_clobber_ref_p_1): Likewise.
10827 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
10828 the alias oracle to disambiguate DRs with stmts DR analysis
10830 (vect_analyze_data_refs): Do not give up on not analyzable
10831 DRs for BB vectorization. Remove code truncating the dataref
10834 2018-05-25 Jakub Jelinek <jakub@redhat.com>
10837 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
10838 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
10839 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
10841 2018-05-25 Richard Biener <rguenther@suse.de>
10843 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
10844 function, combining stmt data ref gathering and fatal analysis
10846 (vect_analyze_data_refs): Remove now redudnant code and simplify.
10847 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
10848 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
10849 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
10850 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
10852 2018-05-25 Bin Cheng <bin.cheng@arm.com>
10854 PR tree-optimization/85720
10855 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
10856 SCC if all partitions are builtins.
10857 (version_loop_by_alias_check): New parameter. Generate cancelable
10858 runtime alias check if all partitions are builtins.
10859 (distribute_loop): Update call to above function.
10861 2018-05-25 Bin Cheng <bin.cheng@arm.com>
10863 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
10864 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
10865 (parm_default_def_partition_arg): Ditto.
10866 (set_parm_default_def_partition): Ditto.
10867 (get_parm_default_def_partitions): Ditto and make it static.
10868 (get_undefined_value_partitions): Ditto and make it static.
10869 (remove_ssa_form): Refactor call to init_var_map here.
10870 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
10871 computation for loop region.
10872 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
10873 (register_default_def): Delete.
10874 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
10875 (parm_default_def_partition_arg): Ditto.
10876 (set_parm_default_def_partition): Ditto.
10877 (get_parm_default_def_partitions): Ditto and make it static.
10878 (get_undefined_value_partitions): Ditto and make it static.
10879 (coalesce_with_default, coalesce_with_default): Update comment.
10880 (create_coalesce_list_for_region): New func factored out from
10881 create_outofssa_var_map.
10882 (populate_coalesce_list_for_outofssa): New func factored out from
10883 create_outofssa_var_map and coalesce_ssa_name.
10884 (create_outofssa_var_map): Delete.
10885 (coalesce_ssa_name): Refactor to support live range computation.
10886 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
10887 (get_parm_default_def_partitions): Delete.
10888 (get_undefined_value_partitions): Ditto.
10889 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
10890 computation for loop region.
10891 (new_tree_live_info, loe_visit_block): Ditto.
10892 (live_worklist, set_var_live_on_entry): Ditto.
10893 (calculate_live_on_exit, verify_live_on_entry): Ditto.
10894 * tree-ssa-live.h (struct _var_map): New fields.
10895 (init_var_map): Change decl.
10896 (region_contains_p): New.
10898 2018-05-25 Bin Cheng <bin.cheng@arm.com>
10900 * tree-ssa-live.h (live_merge_and_clear): Delete.
10902 2018-05-25 Richard Biener <rguenther@suse.de>
10905 * tree-dump.c (dequeue_and_dump): Remove access to removed
10906 operand 2 of a SWITCH_EXPR.
10908 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10910 * doc/sourcebuild.texi (vect_double_cond_arith): Include
10911 multiplication and division.
10912 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
10913 (cond_udiv@var{m}, cond_umod@var{m}): Document.
10914 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
10915 (cond_udiv_optab, cond_umod_optab): New optabs.
10916 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
10917 (IFN_COND_RDIV): New internal functions.
10918 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
10919 TRUNC_MOD_EXPR and RDIV_EXPR.
10920 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
10921 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
10923 (SVE_INT_BINARY): Include mult.
10924 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
10925 (optab, sve_int_op): Handle mult.
10926 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
10928 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
10929 for SVE_INT_BINARY_SD.
10931 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10933 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
10934 (optab, sve_int_op): Handle div and udiv.
10935 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
10936 for SVE_INT_BINARY_SD.
10937 (*<optab><mode>3): New insn for the same.
10939 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10941 * tree-vect-patterns.c: Include predict.h.
10942 (vect_recog_divmod_pattern): Restrict check for division support
10943 to when optimizing for size.
10945 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10947 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
10948 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
10949 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
10950 (gimple_match_op::set_op): Likewise.
10951 (gimple_resimplify4): Declare.
10952 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
10953 (expr::gen_transform): Likewise.
10954 (decision_tree::gen): Generate a simplification routine for 4 operands.
10955 * gimple-match-head.c (gimple_simplify): Add an overload for
10956 4 operands. In the top-level function, handle up to 4 call
10957 arguments and call gimple_resimplify4.
10958 (gimple_resimplify4): New function.
10959 (build_call_internal): Pass a fourth operand.
10960 (maybe_push_to_seq): Likewise.
10961 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
10962 Fold VEC_COND_EXPRs of an operation and a default value into
10963 an IFN_COND_* function if possible.
10964 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
10966 (SVE_COND_FP_BINARY): Include them.
10967 (optab, sve_fp_op): Handle them.
10968 (SVE_INT_BINARY_REV): New code iterator.
10969 (SVE_COND_FP_BINARY_REV): New int iterator.
10970 (commutative): New int attribute.
10971 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
10973 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
10975 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
10976 (*cond_<optab><mode>): New patterns for reversed operands.
10978 2018-05-25 Richard Biener <rguenther@suse.de>
10980 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
10981 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
10982 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
10983 (STMT_VINFO_GROUPED_ACCESS): Adjust.
10984 * tree-vect-data-refs.c (everywhere): Adjust users.
10985 * tree-vect-loop.c (everywhere): Likewise.
10986 * tree-vect-slp.c (everywhere): Likewise.
10987 * tree-vect-stmts.c (everywhere): Likewise.
10988 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
10990 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10992 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
10994 (gcc_cv_as_section_exclude): ... this.
10995 Try Solaris as #exclude syntax.
10996 * configure: Regenerate.
10997 * config.in: Regenerate.
10998 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
11000 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
11001 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
11003 * varasm.c (default_elf_asm_named_section): Don't check if
11004 HAVE_GAS_SECTION_EXCLUDE is defined.
11006 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11008 * doc/md.texi: Update the documentation of the cond_* optabs
11009 to mention the new final operand. Fix GET_MODE_NUNITS call.
11010 Describe the scalar case too.
11011 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
11012 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
11014 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
11015 (get_conditional_internal_fn): Update comment.
11016 * tree-vect-loop.c (vectorizable_reduction): Pass the original
11017 accumulator value as a final argument to conditional functions.
11018 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
11019 a define_expand and add an "else" operand. Assert for now that
11020 the else operand is equal to operand 2. Use SVE_INT_BINARY and
11021 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
11022 (*cond_<optab><mode>): New patterns.
11023 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
11024 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11025 (UNSPEC_COND_EOR): Delete.
11026 (optab): Remove associated mappings.
11027 (SVE_INT_BINARY): New code iterator.
11028 (sve_int_op): Remove int attribute and add "minus" to the code
11030 (SVE_COND_INT_OP): Delete.
11031 (SVE_COND_FP_OP): Rename to...
11032 (SVE_COND_FP_BINARY): ...this.
11034 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11036 * optabs.c (can_reuse_operands_p): New function.
11037 (maybe_legitimize_operands): Try to reuse the results for
11040 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
11042 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
11043 Add {q} suffix to insn mnemonic.
11045 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11047 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
11048 (msp430_warn_func_return): New.
11050 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
11052 * fold-const.c (tree_nonzero_bits): New function.
11053 * fold-const.h (tree_nonzero_bits): Likewise.
11054 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
11055 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
11057 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
11061 * varasm.c (assemble_alias): Check ifunc_resolver only on
11064 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
11067 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
11068 when memory input operand is handled.
11070 2018-05-24 Luis Machado <luis.machado@linaro.org>
11072 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
11074 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
11076 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
11078 * match.pd: Delay FMA folds until after vectorization.
11080 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
11083 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11084 address check not strict.
11086 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
11088 * gimple-match.h (gimple_match_op): New class.
11089 (mprts_hook): Replace parameters with a gimple_match_op *.
11090 (maybe_build_generic_op): Likewise.
11091 (gimple_simplified_result_is_gimple_val): Replace parameters with
11092 a const gimple_match_op *.
11093 (gimple_simplify): Replace code_helper * and tree * parameters with
11094 a gimple_match_op * parameter.
11095 (gimple_resimplify1): Replace code_helper *, tree and tree *
11096 parameters with a gimple_match_op * parameter.
11097 (gimple_resimplify2): Likewise.
11098 (gimple_resimplify3): Likewise.
11099 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
11100 parameters with a gimple_match_op * parameter.
11101 * gimple-match-head.c (gimple_simplify): Change prototypes of
11102 auto-generated functions to take a gimple_match_op * instead of
11103 separate code_helper * and tree * parameters. Make the same
11104 change in the top-level overload and update calls to the
11105 gimple_resimplify routines. Update calls to the auto-generated
11106 functions and to maybe_push_res_to_seq in the publicly-facing
11107 operation-specific gimple_simplify overloads.
11108 (gimple_match_op::MAX_NUM_OPS): Define.
11109 (gimple_resimplify1): Replace rcode and ops with a single res_op
11110 parameter. Update call to gimple_simplify.
11111 (gimple_resimplify2): Likewise.
11112 (gimple_resimplify3): Likewise.
11113 (mprts_hook): Replace parameters with a gimple_match_op *.
11114 (maybe_build_generic_op): Likewise.
11115 (build_call_internal): Replace type, nargs and ops with
11116 a gimple_match_op *.
11117 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
11118 with a single gimple_match_op *. Update calls to mprts_hook,
11119 build_call_internal and gimple_simplified_result_is_gimple_val.
11120 Factor out code that is common to the tree_code and combined_fn cases.
11121 * genmatch.c (expr::gen_transform): Replace tem_code and
11122 tem_ops with a gimple_match_op called tem_op. Update calls
11123 to the gimple_resimplify functions and maybe_push_res_to_seq.
11124 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
11125 res_ops. Update call to the gimple_resimplify functions.
11126 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
11127 (decision_tree::gen): Make the functions take a gimple_match_op *
11128 called res_op instead of separate res_code and res_ops parameters.
11129 Update call accordingly.
11130 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
11131 and ops with a single res_op parameter. Update calls to
11132 maybe_build_generic_op and maybe_push_res_to_seq.
11133 (fold_stmt_1): Update calls to gimple_simplify and
11134 replace_stmt_with_simplification.
11135 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
11136 and gimple_simplified_result_is_gimple_val.
11137 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
11139 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
11140 with a gimple_match_op *.
11141 (vn_nary_build_or_lookup): Likewise. Update call to
11142 vn_nary_build_or_lookup_1.
11143 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
11144 gimple_match_op *. Update calls to the gimple_resimplify routines
11145 and to gimple_simplified_result_is_gimple_val.
11146 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
11147 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
11148 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
11149 (visit_nary_op): Likewise.
11150 (visit_reference_op_load): Likewise.
11152 2018-05-23 Luis Machado <luis.machado@linaro.org>
11154 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
11155 modifier for printing the step amount.
11157 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
11160 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
11163 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
11165 * doc/sourcebuild.texi (Endianness): New subsubsection.
11167 2018-05-23 Luis Machado <luis.machado@linaro.org>
11169 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11170 <prefetch_dynamic_strides>: New const bool field.
11171 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11172 prefetch_dynamic_strides.
11173 (exynosm1_prefetch_tune): Likewise.
11174 (thunderxt88_prefetch_tune): Likewise.
11175 (thunderx_prefetch_tune): Likewise.
11176 (thunderx2t99_prefetch_tune): Likewise.
11177 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
11179 (aarch64_override_options_internal): Update to set
11180 PARAM_PREFETCH_DYNAMIC_STRIDES.
11181 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11182 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11183 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11184 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11185 prefetch-dynamic-strides setting.
11187 2018-05-23 Luis Machado <luis.machado@linaro.org>
11189 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11190 <minimum_stride>: New const int field.
11191 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11192 minimum_stride field defaulting to -1.
11193 (exynosm1_prefetch_tune): Likewise.
11194 (thunderxt88_prefetch_tune): Likewise.
11195 (thunderx_prefetch_tune): Likewise.
11196 (thunderx2t99_prefetch_tune): Likewise.
11197 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
11198 <default_opt_level>: Set to 3.
11199 (aarch64_override_options_internal): Update to set
11200 PARAM_PREFETCH_MINIMUM_STRIDE.
11201 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11202 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11203 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11204 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11205 stride is constant and is below the minimum stride threshold.
11207 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11209 * config/arm/arm-cpus.in (mode26): Delete.
11210 (armv4): Delete mode26 reference.
11211 * config/arm/arm.c (arm_configure_build_target): Delete use of
11214 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
11216 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
11218 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
11219 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
11221 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
11222 (floatunsdidf2): Ditto.
11224 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
11226 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
11227 (fixuns_trunc<mode>si2_avx512f): Ditto.
11228 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
11229 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
11230 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
11232 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
11234 PR rtl-optimization/79985
11235 * df-scan.c (df_insn_refs_collect): Remove special case for
11236 global registers and asm statements.
11238 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
11240 * extend.texi (Global Register Variables): Rewrite the bullet list.
11241 Note that the register is available for allocation. Note that access
11242 via inline asm must use constraints. Add note about async-signal
11243 handlers. Remove paragraph about automagic register selection.
11245 2018-05-23 Richard Biener <rguenther@suse.de>
11247 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
11248 of fixed offset from memset VN.
11250 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
11252 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
11253 first_interp field.
11254 (alloc_cand_and_find_basis): Initialize first_interp field.
11255 (slsr_process_mul): Modify first_interp field.
11256 (slsr_process_add): Likewise.
11257 (slsr_process_cast): Modify first_interp field for each new
11259 (slsr_process_copy): Likewise.
11260 (dump_candidate): Dump first_interp field.
11261 (replace_mult_candidate): Process all interpretations, not just
11263 (replace_rhs_if_not_dup): Likewise.
11264 (replace_one_candidate): Likewise.
11266 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
11268 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
11270 (aarch64_needs_frame_chain): New function.
11271 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
11273 2018-05-23 Sudakshina Das <sudi.das@arm.com>
11276 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
11277 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
11278 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
11279 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
11280 as true for strict-align.
11281 (aarch64_can_inline_p): Perform checks even when callee has no
11282 attributes to check for strict alignment.
11283 * doc/extend.texi (AArch64 Function Attributes): Document
11285 * doc/invoke.texi: (AArch64 Options): Likewise.
11287 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
11289 PR tree-optimization/85853
11290 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
11291 the handling of the root of the node to...
11292 (vect_slp_analyze_node_operations_1): ...this new function,
11293 and run the whole thing with the child nodes' def types
11294 set according to their SLP node's def type.
11296 2018-05-23 Richard Biener <rguenther@suse.de>
11298 PR middle-end/85874
11299 * tree-data-ref.c (create_runtime_alias_checks): Defer
11300 and ignore overflow warnings.
11302 2018-05-23 Yury Gribov <tetra2005@gmail.com>
11304 PR tree-optimization/85822
11305 * tree-vrp.c (is_masked_range_test): Fix handling of negative
11308 2018-05-23 Richard Biener <rguenther@suse.de>
11310 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
11311 memset constants via native_interpret_expr.
11313 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
11316 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
11318 (cgraph_node::create_alias): Likewise.
11319 (cgraph_node::get_availability): Check ifunc_resolver instead
11320 of looking up ifunc attribute.
11321 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
11322 * varasm.c (do_assemble_alias): Likewise.
11323 (assemble_alias): Likewise.
11324 (default_binds_local_p_3): Likewise.
11325 * cgraph.h (cgraph_node): Add ifunc_resolver.
11326 (cgraph_node::only_called_directly_or_aliased_p): Return false
11327 for IFUNC resolver.
11328 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
11330 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
11331 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
11332 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
11333 instead of looking up ifunc attribute.
11335 2018-05-22 Luis Machado <luis.machado@linaro.org>
11337 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
11339 2018-05-22 Martin Sebor <msebor@redhat.com>
11341 PR middle-end/85359
11342 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
11343 only when expasion succeeds.
11344 (expand_builtin_strcmp): Same.
11345 (expand_builtin_strncmp): Same.
11347 2018-05-22 Martin Sebor <msebor@redhat.com>
11349 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
11351 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
11352 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11354 * config/aarch64/aarch64-ldpstp.md: Replace uses of
11355 aarch64_mem_pair_operand with memory_operand and delete operand swapping
11357 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
11358 Add check for legitimate_address.
11359 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
11360 (aarch64_swap_ldrstr_operands): New.
11361 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
11364 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
11365 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11367 * config/aarch64/aarch64.md: New patterns to generate stp
11369 (store_pair_sw, store_pair_dw): New patterns to generate stp for
11370 single words and double words.
11371 (load_pair_sw, load_pair_dw): Likewise.
11372 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
11374 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
11376 * config/aarch64/aarch64-ldpstp.md: Modify peephole
11377 for different mode ldpstp and add peephole for merged zero stores.
11378 Likewise for loads.
11379 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
11381 (aarch64_gen_store_pair): Rename calls to match new patterns.
11382 (aarch64_gen_load_pair): Rename calls to match new patterns.
11383 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
11384 (load_pair<DREG:mode><DREG2:mode>): ... This.
11385 (store_pair<mode>): Rename to...
11386 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
11387 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
11388 New mode iterators.
11389 (V_INT_EQUIV): Handle SImode.
11390 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
11393 2018-05-22 Martin Sebor <msebor@redhat.com>
11396 * calls.c (maybe_warn_nonstring_arg): Use string length to set
11397 or ajust the presumed bound on an operation to avoid unnecessary
11400 2018-05-22 Martin Sebor <msebor@redhat.com>
11402 PR tree-optimization/85826
11403 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
11404 assuming that a DECL necesarily has a constant size.
11406 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
11408 PR middle-end/85862
11409 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
11411 2018-05-22 Richard Biener <rguenther@suse.de>
11413 PR tree-optimization/85834
11414 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
11415 non-constant and non-zero memset arguments.
11417 2018-05-22 Martin Liska <mliska@suse.cz>
11420 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
11422 2018-05-22 Richard Biener <rguenther@suse.de>
11424 PR tree-optimization/85863
11425 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
11426 comparisons when vectype is specified.
11427 (vectorizable_condition): Do not specify vectype for
11428 vect_is_simple_cond when SLP vectorizing.
11430 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
11433 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
11434 define __ibm128 as long double.
11435 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
11436 as a distinct type when IEEE 128-bit support is enabled.
11437 (init_float128_ieee): Fix up conversions between IFmode and IEEE
11438 128-bit types to use the correct functions.
11439 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
11440 convert between 128-bit floating point types that have different
11441 modes but the same representation, instead of using gen_lowpart to
11443 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
11445 (IFKF_reg): New attributes to give the register constraints for
11447 (extend<mode>tf2_internal): New insns to mark an explicit
11448 conversion between 128-bit floating point types that have a
11449 different mode but share the same representation.
11451 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
11453 PR tree-optimization/85814
11454 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
11455 a null return from get_strinfo when unsharing the next
11456 strinfo in the chain.
11458 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
11461 * varasm.c (weak_finish): Clean up weak_decls.
11463 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11465 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
11466 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
11467 UNSPEC_UADALP values.
11468 * config/aarch64/iterators.md (ABAL): New int iterator.
11471 (sur): Add mappings for the above.
11472 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
11474 (aarch64_<sur>abal<mode>_4): Likewise.
11475 (aarch64_<sur>adalp<mode>_3): Likewise.
11476 (<sur>sadv16qi): New define_expand.
11478 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
11480 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
11481 (*movdf_internal): Ditto.
11482 (*rcpsf2_sse): Ditto.
11483 (*rsqrtsf2_sse): Ditto.
11484 (*sqrt<mode>2_sse): Ditto.
11486 2018-05-21 Tamar Christina <tamar.christina@arm.com>
11488 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
11490 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
11491 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
11492 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
11493 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
11495 * config/aarch64/arm_neon.h: Likewise.
11496 * config/aarch64/iterators.md (VQ_I): New.
11498 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
11500 * config.gcc: Add arc/t-multilib-linux to tmake_file for
11502 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
11505 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
11507 * config/nds32/constraints.md (S): New constraint.
11508 * config/nds32/nds32.md (call_internal): Use constraint S.
11509 (call_value_internal): Likewise.
11510 (sibcall_internal): Likewise.
11511 (sibcall_value_internal): Likewise.
11513 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
11514 Chung-Ju Wu <jasonwucj@gmail.com>
11516 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
11517 into consideration.
11519 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
11520 Chung-Ju Wu <jasonwucj@gmail.com>
11522 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
11523 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
11524 (nds32_rtx_costs_impl): Simplify.
11525 (nds32_address_cost_impl): Simplify.
11526 (nds32_init_rtx_costs): New function.
11527 (nds32_rtx_costs_speed_prefer): Likewise.
11528 (nds32_rtx_costs_size_prefer): Likewise.
11529 (nds32_address_cost_speed_prefer): Likewise.
11530 (nds32_address_cost_speed_fwprop): Likewise.
11531 (nds32_address_cost_size_prefer): Likewise.
11532 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
11533 * config/nds32/nds32.c (nds32_option_override): Use
11534 nds32_init_rtx_costs function.
11536 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
11538 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
11539 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
11540 (TARGET_PIPELINE_N8): Likewise.
11541 (TARGET_PIPELINE_N10): Likewise.
11542 (TARGET_PIPELINE_N13): Likewise.
11543 (TARGET_PIPELINE_GRAYWOLF): Likewise.
11545 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
11547 * config/nds32/nds32-fpu.md: Update copyright year.
11549 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11551 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
11553 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11555 * config/nds32/nds32.c
11556 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
11557 * config/nds32/nds32.opt (minline-asm-r15): New option.
11559 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11561 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
11563 * config/nds32/nds32.md (abssi2): New pattern.
11565 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
11567 * config/i386/i386.md (rex64namesuffix): New mode attribute.
11568 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
11569 Merge insn pattern from sse_cvtsi2ss<round_name> and
11570 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
11571 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
11572 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
11573 using SWI48 mode iterator.
11574 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
11575 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
11576 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
11577 pattern from sse_cvttss2si<round_saeonly_name>
11578 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
11579 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
11580 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
11581 using SWI48 mode iterator.
11582 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
11583 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
11584 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
11585 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
11586 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
11587 using SWI48 mode iterator.
11588 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
11589 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
11590 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
11591 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
11592 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
11593 SWI48 mode iterator.
11594 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
11595 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
11596 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
11597 pattern from sse_cvttsd2si<round_saeonly_name>
11598 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
11600 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11602 * config/nds32/nds32-md-auxiliary.c
11603 (nds32_valid_smw_lwm_base_p): Refine.
11604 (nds32_output_smw_single_word): Refine.
11605 (nds32_output_smw_double_word): New.
11606 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
11608 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11610 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
11611 (nds32_output_stack_pop): Refine.
11612 (nds32_expand_unaligned_load): Refine.
11613 (nds32_expand_unaligned_store): Refine.
11615 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11616 Chung-Ju Wu <jasonwucj@gmail.com>
11618 * config/nds32/constants.md: Add TP_REGNUM constant.
11619 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
11620 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
11622 * config/nds32/nds32-doubleword.md: Consider flag_pic.
11623 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
11624 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
11625 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
11626 and PIC code generation.
11627 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
11629 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
11631 * config/nds32/nds32.md: Support TLS and PIC.
11632 * config/nds32/nds32.c: Support TLS and PIC.
11633 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
11634 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
11637 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11639 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
11640 mode with E_ prefix.
11642 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11643 Chung-Ju Wu <jasonwucj@gmail.com>
11645 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
11646 * config/nds32/nds32-md-auxiliary.c
11647 (symbolic_reference_mentioned_p): New.
11648 (nds32_legitimize_ict_address): New.
11649 (nds32_expand_ict_move): New.
11650 (nds32_indirect_call_referenced_p): New.
11651 (nds32_symbol_binds_local_p): Delete.
11652 (nds32_long_call_p): Modify.
11653 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
11654 * config/nds32/nds32-protos.h
11655 (symbolic_reference_mentioned_p): Declare.
11656 (nds32_legitimize_ict_address): Declare.
11657 (nds32_expand_ict_move): Declare.
11658 (nds32_indirect_call_referenced_p): Declare.
11659 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
11660 (nds32_relax_group): Use nds32_ict_const_p as condition.
11661 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
11662 (nds32_asm_file_start): Output ict_model directive in asm code.
11663 (nds32_legitimate_address_p): Consider indirect call.
11664 (nds32_print_operand): Consider indirect call.
11665 (nds32_print_operand_address): Consider indirect call.
11666 (nds32_insert_attributes): Handle "indirect_call" attribute.
11667 (TARGET_LEGITIMATE_ADDRESS_P): Define.
11668 (TARGET_LEGITIMATE_CONSTANT_P): Define.
11669 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
11670 (TARGET_DELEGITIMIZE_ADDRESS): Define.
11671 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
11672 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
11673 (TARGET_ICT_MODEL_SMALL): Define.
11674 (TARGET_ICT_MODEL_LARGE): Define.
11675 * config/nds32/nds32.md (movsi): Consider ict model.
11676 (call, call_value): Consider ict model.
11677 (sibcall, sibcall_value): Consider ict model.
11678 * config/nds32/nds32.opt (mict-model): New option.
11679 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
11682 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
11683 Monk Chiang <sh.chiang04@gmail.com>
11684 Jim Wilson <jimw@sifive.com>
11686 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
11687 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
11688 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
11689 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
11690 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
11691 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
11692 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
11693 compute save_libcall_adjustment properly.
11694 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
11695 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
11696 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
11697 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
11698 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
11699 (ABI_SPEC): Handle mabi=ilp32e.
11700 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
11701 (RVE): Add RVE mask.
11702 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
11703 <-march>: Add rv32e as an example.
11705 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
11708 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
11709 (intra_create_variable_infos): Handle C++ constructors.
11711 2018-05-18 Martin Liska <mliska@suse.cz>
11713 * passes.def: Remove a redundant pass.
11715 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
11718 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
11720 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11722 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
11724 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
11726 (armv2, armv2a, armv3, armv3m): Delete architectures.
11727 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
11728 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
11729 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
11731 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
11732 (*mulsidi3adddi): Likewise.
11733 (mulsidi3): Likewise.
11734 (*mulsidi3_nov6): Likewise.
11735 (umulsidi3): Likewise.
11736 (umulsidi3_nov6): Likewise.
11737 (umaddsidi4): Likewise.
11738 (*umulsidi3adddi): Likewise.
11739 (smulsi3_highpart): Likewise.
11740 (*smulsi3_highpart_nov6): Likewise.
11741 (umulsi3_highpart): Likewise.
11742 (*umulsi3_highpart_nov6): Likewise.
11743 * config/arm/arm.h (arm_arch3m): Delete.
11744 * config/arm/arm.c (arm_arch3m): Delete.
11745 (arm_option_override_internal): Update armv3-related comment.
11746 (arm_configure_build_target): Delete use of isa_bit_mode32.
11747 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
11748 (arm_rtx_costs_internal): Delete check of arm_arch3m.
11749 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
11750 (mulsa3): Likewise.
11751 (mulusa3): Likewise.
11752 * config/arm/arm-protos.h (arm_arch3m): Delete.
11753 * config/arm/arm-tables.opt: Regenerate.
11754 * config/arm/arm-tune.md: Likewise.
11755 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
11756 deleted architectures.
11758 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11760 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
11761 (armv5t, armv5te): New features.
11762 (ARMv5, ARMv5e): Delete fgroups.
11763 (ARMv5t, ARMv5te): Adjust for above changes.
11764 (ARMv6m): Likewise.
11765 (armv5, armv5e): Delete arches.
11766 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
11768 (*call_reg_arm): Likewise.
11769 (*call_value_reg_armv5): Likewise.
11770 (*call_value_reg_arm): Likewise.
11771 (*call_symbol): Likewise.
11772 (*call_value_symbol): Likewise.
11773 (*sibcall_insn): Likewise.
11774 (*sibcall_value_insn): Likewise.
11775 (clzsi2): Likewise.
11776 (prefetch): Likewise.
11777 (define_split and define_peephole2 dependent on arm_arch5):
11779 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
11781 (TARGET_ARM_QBIT): Likewise.
11782 (TARGET_DSP_MULTIPLY): Likewise.
11783 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
11784 (arm_arch5, arm_arch5e): Delete.
11785 (arm_arch5t, arm_arch5te): Declare.
11786 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
11787 (arm_arch5t): Declare.
11788 (arm_option_reconfigure_globals): Update for the above.
11789 (arm_options_perform_arch_sanity_checks): Update comment, replace
11790 use of arm_arch5 with arm_arch5t.
11791 (use_return_insn): Likewise.
11792 (arm_emit_call_insn): Likewise.
11793 (output_return_instruction): Likewise.
11794 (arm_final_prescan_insn): Likewise.
11795 (arm_coproc_builtin_available): Likewise.
11796 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
11797 arm_arch5e with arm_arch5t and arm_arch5te.
11798 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
11799 (arm_arch5t, arm_arch5te): Declare.
11800 * config/arm/arm-tables.opt: Regenerate.
11801 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
11802 * config/arm/t-multilib: Likewise.
11803 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
11804 instead of arm_arch5.
11805 (*call_reg_thumb1): Likewise.
11806 (*call_value_reg_thumb1_v5): Likewise.
11807 (*call_value_reg_thumb1): Likewise.
11808 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
11810 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
11812 2018-05-18 Martin Liska <mliska@suse.cz>
11814 PR gcov-profile/84846
11815 * doc/gcov.texi: Document -t option of gcov tool.
11817 2018-05-18 Martin Liska <mliska@suse.cz>
11819 PR gcov-profile/84846
11820 * gcov.c (print_usage): Add new -t option.
11821 (process_args): Handle the option.
11822 (generate_results): Use stdout as output when requested by
11825 2018-05-18 Martin Liska <mliska@suse.cz>
11827 PR gcov-profile/84846
11828 * coverage.c (coverage_init): Write PWD to .gcno file.
11829 * doc/gcov.texi: Document how working directory is printed.
11830 * gcov-dump.c (dump_gcov_file): Print PWD.
11831 * gcov.c (output_intermediate_file): Likewise.
11832 (read_graph_file): Read PWD string.
11833 (output_lines): Print PWD.
11835 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11837 PR middle-end/85817
11838 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
11839 for retval and return false if all args to phi are zero.
11841 2018-05-18 Richard Biener <rguenther@suse.de>
11843 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
11845 (evrp_dom_walker::before_dom_children): Call it.
11847 2018-05-18 Richard Biener <rguenther@suse.de>
11849 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
11850 results when processing array refs with variable index.
11852 2018-05-18 Toon Moene <toon@moene.org>
11854 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
11855 directly after that of -floop-interchange. Indicate that both
11856 options are enabled by default when specifying -O3.
11858 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11860 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
11861 iterator. Delete separate integer-mode vec_set<mode> expander.
11862 (aarch64_simd_vec_setv2di): Delete.
11863 (vec_setv2di): Delete.
11864 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
11865 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
11866 the "w, r" alternative.
11868 2018-05-18 Martin Liska <mliska@suse.cz>
11870 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
11871 * tree-pass.h (make_pass_lower_switch_O0): New function.
11872 * tree-switch-conversion.c (node_has_low_bound): Remove.
11873 (node_has_high_bound): Likewise.
11874 (node_is_bounded): Likewise.
11875 (class pass_lower_switch): Make it a template type and create
11877 (pass_lower_switch::execute): Add template argument.
11878 (make_pass_lower_switch): New function.
11879 (make_pass_lower_switch_O0): New function.
11880 (do_jump_if_equal): Remove.
11881 (emit_case_nodes): Simplify to just handle all 3 cases and leave
11882 all the hard work to tree optimization passes.
11884 2018-05-18 Martin Liska <mliska@suse.cz>
11886 * dbgcnt.c (limit_low): Renamed from limit.
11887 (limit_high): New variable.
11888 (dbg_cnt_is_enabled): Check for upper limit.
11889 (dbg_cnt): Adjust dumping.
11890 (dbg_cnt_set_limit_by_index): Add new argument for high
11892 (dbg_cnt_set_limit_by_name): Likewise.
11893 (dbg_cnt_process_single_pair): Parse new format.
11894 (dbg_cnt_process_opt): Use strtok.
11895 (dbg_cnt_list_all_counters): Remove 'value' and add
11897 * doc/invoke.texi: Document changes.
11899 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
11901 * doc/sourcebuild.texi (scalar_all_fma): Document.
11902 * tree.def (FMA_EXPR): Delete.
11903 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
11904 * internal-fn.c (ternary_direct): New macro.
11905 (expand_ternary_optab_fn): Likewise.
11906 (direct_ternary_optab_supported_p): Likewise.
11907 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
11908 * builtins.c (fold_builtin_fma): Delete.
11909 (fold_builtin_3): Don't call it.
11910 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
11911 * expr.c (expand_expr_real_2): Likewise.
11912 * fold-const.c (operand_equal_p): Likewise.
11913 (fold_ternary_loc): Likewise.
11914 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
11915 * gimple.c (DEFTREECODE): Likewise.
11916 * gimplify.c (gimplify_expr): Likewise.
11917 * optabs-tree.c (optab_for_tree_code): Likewise.
11918 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
11919 * tree-eh.c (operation_could_trap_p): Likewise.
11920 (stmt_could_throw_1_p): Likewise.
11921 * tree-inline.c (estimate_operator_cost): Likewise.
11922 * tree-pretty-print.c (dump_generic_node): Likewise.
11923 (op_code_prio): Likewise.
11924 * tree-ssa-loop-im.c (stmt_cost): Likewise.
11925 * tree-ssa-operands.c (get_expr_operands): Likewise.
11926 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
11927 * fold-const-call.h (fold_fma): Delete.
11928 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
11929 CFN_FNMA and CFN_FNMS.
11930 (fold_fma): Delete.
11931 * genmatch.c (combined_fn): New enum.
11932 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
11933 (commutative_op): New function.
11934 (commutate): Use it. Handle more than 2 operands.
11935 (dt_operand::gen_gimple_expr): Use commutative_op.
11936 (parser::parse_expr): Allow :c to be used with non-binary
11937 operators if the commutative operand is known.
11938 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
11939 CFN_FMS, CFN_FNMA and CFN_FNMS.
11940 (backprop::process_assign_use): Remove FMA_EXPR handling.
11941 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
11942 (gen_hsa_fma): New function.
11943 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
11944 IFN_FNMA and IFN_FNMS.
11945 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
11946 * gimple-fold.h (follow_all_ssa_edges): Declare.
11947 * gimple-fold.c (follow_all_ssa_edges): New function.
11948 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
11949 gimple_build interface and use follow_all_ssa_edges to fold the result.
11950 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
11951 instead of checking for optabs directly.
11952 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
11953 rather than FMA_EXPRs.
11954 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
11955 call to IFN_FMA instead of an FMA_EXPR.
11957 2018-05-17 Jim Wilson <jimw@sifive.com>
11959 * expr.c (do_tablejump): When converting index to Pmode, if we have a
11960 sign extended promoted subreg, and the range does not have the sign bit
11961 set, then do a sign extend.
11963 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
11964 test, check for sign extended subreg and/or constant operands, and
11965 do a sign extend in that case.
11967 2018-05-17 Steve Ellcey <sellcey@cavium.com>
11969 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
11970 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
11972 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
11973 Change logics_shift_reg to logics_shift_imm.
11974 (thunderx2t99_fp_loadpair_basic): Delete.
11975 (thunderx2t99_fp_storepair_basic): Delete.
11976 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
11977 (thunderx2t99_asimd_polynomial): Delete.
11978 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
11979 and neon_fp_mul_d_scalar_q.
11980 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
11981 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
11982 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
11983 (thunderx2t99_asimd_lut): Add missing tbl types.
11984 (thunderx2t99_asimd_ext): Delete.
11985 (thunderx2t99_asimd_load1_1_mult): Delete.
11986 (thunderx2t99_asimd_load1_2_mult): Delete.
11987 (thunderx2t99_asimd_load1_ldp): New.
11988 (thunderx2t99_asimd_load1): New.
11989 (thunderx2t99_asimd_load2): Add missing *load2* types.
11990 (thunderx2t99_asimd_load3): New.
11991 (thunderx2t99_asimd_load4): New.
11992 (thunderx2t99_asimd_store1_1_mult): Delete.
11993 (thunderx2t99_asimd_store1_2_mult): Delete.
11994 (thunderx2t99_asimd_store2_mult): Delete.
11995 (thunderx2t99_asimd_store2_onelane): Delete.
11996 (thunderx2t99_asimd_store_stp): New.
11997 (thunderx2t99_asimd_store1): New.
11998 (thunderx2t99_asimd_store2): New.
11999 (thunderx2t99_asimd_store3): New.
12000 (thunderx2t99_asimd_store4): New.
12002 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
12004 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
12005 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
12007 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
12008 Segher Boessenkool <segher@kernel.crashing.org>
12011 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
12014 2018-05-17 Richard Biener <rguenther@suse.de>
12016 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
12017 for pruning loop and prune defs feeding only already visited PHIs.
12019 2018-05-17 Richard Biener <rguenther@suse.de>
12021 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
12023 2018-05-17 Bin Cheng <bin.cheng@arm.com>
12024 Richard Biener <rguenther@suse.de>
12026 PR tree-optimization/85793
12027 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
12028 for VMAT_ELEMENTWISE.
12030 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
12032 * internal-fn.h (lookup_internal_fn): Declare
12033 * internal-fn.c (lookup_internal_fn): New function.
12034 * gimple.c (gimple_build_call_from_tree): Handle calls to
12035 internal functions.
12036 * gimple-pretty-print.c (dump_gimple_call): Print "." before
12037 internal function names.
12038 * tree-pretty-print.c (dump_generic_node): Likewise.
12039 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
12041 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
12043 * gimple-fold.h (gimple_build): Make the function forms take
12044 combined_fn rather than built_in_function.
12045 (gimple_simplify): Likewise.
12046 * gimple-match-head.c (gimple_simplify): Likewise.
12047 * gimple-fold.c (gimple_build): Likewise.
12048 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
12049 rather than gimple_build_call_internal.
12050 (get_initial_defs_for_reduction): Likewise.
12051 (vect_create_epilog_for_reduction): Likewise.
12052 (vectorizable_live_operation): Likewise.
12054 2018-05-17 Martin Liska <mliska@suse.cz>
12056 * gimple-ssa-sprintf.c (format_directive): Do not use
12057 space in between 'G_' and '('.
12059 2018-05-17 Jakub Jelinek <jakub@redhat.com>
12062 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
12063 even if the mask is not all ones.
12066 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
12068 (ix86_gimple_fold_builtin): Likewise.
12071 * config/i386/i386.c: Include tree-vector-builder.h.
12072 (ix86_vector_shift_count): New function.
12073 (ix86_fold_builtin): Fold shift builtins by scalar count.
12074 (ix86_gimple_fold_builtin): Likewise.
12076 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
12077 _mm512_setzero): New intrinsics.
12079 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
12080 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12082 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
12083 code generation for cases where splatting a value is not useful.
12084 * simplify-rtx.c (simplify_ternary_operation): Simplify
12085 vec_merge across a vec_duplicate and a paradoxical subreg forming
12086 a vector mode to a vec_concat.
12088 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
12090 * config.gcc: Support "goldmont-plus".
12091 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
12093 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12094 PROCESSOR_GOLDMONT_PLUS.
12095 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
12096 (processor_target_table): Add "goldmont-plus".
12097 (PTA_GOLDMONT_PLUS): Define.
12098 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
12099 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
12100 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
12101 (fold_builtin_cpu): Add "goldmont-plus".
12102 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
12103 (ix86_option_override_internal): Add "goldmont-plus".
12104 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
12105 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
12106 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
12107 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
12109 2018-05-17 Richard Biener <rguenther@suse.de>
12111 PR tree-optimization/85757
12112 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
12113 remove defs that only feed that PHI from further processing.
12115 2018-05-16 Jim Wilson <jimw@sifive.com>
12117 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
12119 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
12121 2018-05-16 Mark Wielaard <mark@klomp.org>
12123 * dwarf2out.c (count_index_strings): New function.
12124 (output_indirect_strings): Call count_index_strings and generate
12125 header for dwarf_version >= 5.
12127 2018-05-16 Mark Wielaard <mark@klomp.org>
12129 * dwarf2out.c (dwarf_FORM): New function.
12130 (set_indirect_string): Use dwarf_FORM.
12131 (reset_indirect_string): Likewise.
12132 (size_of_die): Likewise.
12133 (value_format): Likewise.
12134 (output_die): Likewise.
12135 (add_skeleton_AT_string): Likewise.
12136 (output_macinfo_op): Likewise.
12137 (index_string): Likewise.
12138 (output_index_string_offset): Likewise.
12139 (output_index_string): Likewise.
12140 (count_index_strings): Likewise.
12142 2018-05-16 Carl Love <cel@us.ibm.com>
12144 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
12145 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
12147 2018-05-16 Martin Jambor <mjambor@suse.cz>
12149 * ipa-prop.c (ipa_free_all_edge_args): Remove.
12150 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
12152 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
12154 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
12155 (fnma<mode>4): Likewise.
12156 (fms<mode>4): Likewise.
12157 (fnms<mode>4): Likewise.
12158 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
12159 (aarch64_fnma<mode>4): Likewise.
12160 (aarch64_fms<mode>4): Likewise.
12161 (aarch64_fnms<mode>4): Likewise.
12162 (aarch64_fnmadd<mode>4): Likewise.
12164 2018-05-16 Jason Merrill <jason@redhat.com>
12166 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
12168 2018-05-16 Richard Biener <rguenther@suse.de>
12170 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
12171 (dump_stmt_cost): Declare.
12172 (add_stmt_cost): Dump cost we add.
12173 (add_stmt_costs): New function.
12174 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
12175 No longer exported.
12176 (vect_analyze_stmt): Adjust prototype.
12177 (vectorizable_condition): Likewise.
12178 (vectorizable_live_operation): Likewise.
12179 (vectorizable_reduction): Likewise.
12180 (vectorizable_induction): Likewise.
12181 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
12182 cost vector to pass to vectorizable_ and record afterwards.
12183 (vect_model_reduction_cost): Take cost vector argument and adjust.
12184 (vect_model_induction_cost): Likewise.
12185 (vectorizable_reduction): Likewise.
12186 (vectorizable_induction): Likewise.
12187 (vectorizable_live_operation): Likewise.
12188 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
12189 SLP_TREE_NUMBER_OF_VEC_STMTS.
12190 (vect_analyze_slp_cost_1): Remove.
12191 (vect_analyze_slp_cost): Likewise.
12192 (vect_slp_analyze_node_operations): Take visited args and
12193 a target cost vector. Avoid processing already visited stmt sets.
12194 (vect_slp_analyze_operations): Use a local cost vector to gather
12195 costs and register those of non-discarded instances.
12196 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
12197 (vect_schedule_slp_instance): Remove copying of
12198 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
12200 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
12201 adding cost. Record cost entry location.
12202 (vect_prologue_cost_for_slp_op): Function to compute cost of
12203 a constant or invariant generated for SLP vect in the prologue,
12204 split out from vect_analyze_slp_cost_1.
12205 (vect_model_simple_cost): Make static. Adjust for SLP costing.
12206 (vect_model_promotion_demotion_cost): Likewise.
12207 (vect_model_store_cost): Likewise, make static.
12208 (vect_model_load_cost): Likewise.
12209 (vectorizable_bswap): Add cost vector arg and adjust.
12210 (vectorizable_call): Likewise.
12211 (vectorizable_simd_clone_call): Likewise.
12212 (vectorizable_conversion): Likewise.
12213 (vectorizable_assignment): Likewise.
12214 (vectorizable_shift): Likewise.
12215 (vectorizable_operation): Likewise.
12216 (vectorizable_store): Likewise.
12217 (vectorizable_load): Likewise.
12218 (vectorizable_condition): Likewise.
12219 (vectorizable_comparison): Likewise.
12220 (can_vectorize_live_stmts): Likewise.
12221 (vect_analyze_stmt): Likewise.
12222 (vect_transform_stmt): Adjust calls to vectorizable_*.
12223 * tree-vectorizer.c: Include gimple-pretty-print.h.
12224 (dump_stmt_cost): New function.
12226 2018-05-16 Richard Biener <rguenther@suse.de>
12228 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
12229 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
12230 * tree-ssa-dse.c: Include tree-ssa-loop.h.
12231 (check_name): New callback.
12232 (dse_classify_store): Track cycles via a visited bitmap of PHI
12233 defs and simplify handling of in-loop and across loop dead stores
12234 and properly fail for loop-variant refs. Handle byte-tracking with
12235 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
12238 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
12240 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
12241 (vect_get_mask_type_for_stmt): Likewise.
12242 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
12244 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
12245 to determine the statement's vector type and the vector type that
12246 should be used for calculating nunits. Deal with cases in which
12247 the type has to be deferred.
12248 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
12249 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
12250 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
12251 (vect_determine_vf_for_stmt): New functions, split out from...
12252 (vect_determine_vectorization_factor): ...here.
12253 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
12254 (vect_get_mask_type_for_stmt): New functions, split out from
12255 vect_determine_vectorization_factor.
12257 2018-05-16 Richard Biener <rguenther@suse.de>
12259 * tree-cfg.c (verify_gimple_assign_ternary): Properly
12260 verify the [VEC_]COND_EXPR embedded comparison.
12262 2018-05-15 Martin Sebor <msebor@redhat.com>
12264 PR tree-optimization/85753
12265 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
12266 RECORD_TYPE in addition to ARRAY_TYPE.
12268 2018-05-15 Martin Sebor <msebor@redhat.com>
12270 PR middle-end/85643
12271 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
12273 2018-05-15 Richard Biener <rguenther@suse.de>
12275 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
12276 add by_clobber_p one. Change algorithm to collect all defs
12277 representing uses we need to walk and try reducing them to
12278 a single one before failing.
12279 (dse_dom_walker::dse_optimize_stmt): Adjust.
12281 2018-05-13 Mark Wielaard <mark@klomp.org>
12283 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
12284 (size_of_loc_descr): Likewise.
12285 (output_loc_operands): Likewise.
12286 (output_loc_operands_raw): Likewise.
12287 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
12288 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
12289 (hash_loc_operands): Likewise.
12290 (compare_loc_operands): Likewise.
12292 2018-05-14 Mark Wielaard <mark@klomp.org>
12294 * dwarf2out.c (count_index_addrs): New function.
12295 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
12297 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12299 PR tree-optimization/83648
12300 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
12301 return value as malloc candidate.
12303 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12306 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
12307 param as true in call to suggest_attribute.
12309 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
12311 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
12312 -mreadonly-in-sdata.
12314 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12316 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
12318 (aarch64_crypto_aesd_fused): Likewise.
12320 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
12322 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
12323 (movsi_aarch64): Likewise.
12324 (load_pairsi): Likewise.
12325 (load_pairdi): Likewise.
12326 (store_pairsi): Likewise.
12327 (store_pairdi): Likewise.
12328 (load_pairsf): Likewise.
12329 (load_pairdf): Likewise.
12330 (store_pairsf): Likewise.
12331 (store_pairdf): Likewise.
12332 (zero_extend): Likewise.
12333 (trunc): Swap alternatives.
12334 (fcvt_target): Add '?' to prefer w over r.
12336 2018-05-14 Jakub Jelinek <jakub@redhat.com>
12339 * config/i386/i386.md: Disallow non-commutative arithmetics in
12340 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
12341 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
12342 in the peephole2 before it.
12344 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
12346 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
12347 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
12348 (ix86_handle_option): Handle -mcldemote.
12349 * config.gcc: New header.
12350 * config/i386/cldemoteintrin.h: New file.
12351 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
12352 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
12354 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12355 OPTION_MASK_ISA_CLDEMOTE.
12356 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
12357 (ix86_valid_target_attribute_inner_p): Ditto.
12358 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
12359 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
12360 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
12361 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
12362 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
12364 * config/i386/i386.opt: Add -mcldemote.
12365 * config/i386/x86intrin.h: New header.
12366 * doc/invoke.texi: Add -mcldemote.
12368 2018-05-14 Richard Biener <rguenther@suse.de>
12370 * doc/match-and-simplify.texi: Adjust :s documentation.
12372 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
12374 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
12375 intended memcpy size.
12376 (REORDER_45): Likewise.
12378 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
12380 * sort.cc: New file.
12381 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
12382 * vec.c (qsort_chk): Use gcc_qsort.
12383 * Makefile.in (OBJS-libcommon): Add sort.o.
12384 (build/sort.o): New target. Use it...
12385 (BUILD_RTL): ... here, and...
12386 (build/gencfn-macros): ... here, and...
12387 (build/genmatch): ... here.
12389 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
12390 Chung-Ju Wu <jasonwucj@gmail.com>
12392 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
12393 * config/nds32/nds32-graywolf.md: New file.
12394 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
12395 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
12397 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
12398 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
12399 * config/nds32/nds32.md (pipeline_model): Add graywolf.
12400 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
12401 * config/nds32/pipelines.md: Include n15 settings.
12403 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
12404 Chung-Ju Wu <jasonwucj@gmail.com>
12406 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
12407 * config/nds32/nds32-n13.md: New file.
12408 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
12409 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
12411 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
12412 * config/nds32/nds32.md (pipeline_model): Add n13.
12413 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
12414 * config/nds32/pipelines.md: Include n13 settings.
12416 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
12417 Chung-Ju Wu <jasonwucj@gmail.com>
12419 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
12420 * config/nds32/nds32-n10.md: New file.
12421 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
12422 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
12424 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
12425 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
12426 * config/nds32/nds32.md (pipeline_model): Add n10.
12427 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
12428 * config/nds32/pipelines.md: Include n10 settings.
12430 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
12431 Kito Cheng <kito.cheng@gmail.com>
12432 Chung-Ju Wu <jasonwucj@gmail.com>
12434 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
12435 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
12436 Add enum values for DSP extension instructions.
12437 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
12439 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
12440 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
12441 New code iterators.
12442 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
12443 * config/nds32/nds32-dspext.md: New file for DSP implementation.
12444 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
12445 * config/nds32/nds32-intrinsic.md: Likewise.
12446 * config/nds32/nds32_intrinsic.h: Likewise.
12447 * config/nds32/nds32-md-auxiliary.c: Likewise.
12448 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
12449 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
12450 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
12451 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
12452 * config/nds32/nds32-protos.h: New declarations for DSP extension.
12453 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
12454 TYPE_DMAC in switch statement.
12455 * config/nds32/nds32.c: New checking and implementation for DSP
12456 extension instructions.
12457 * config/nds32/nds32.h: Likewise.
12458 * config/nds32/nds32.md: Likewise.
12459 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
12460 * config/nds32/predicates.md: Implement new predicates for DSP
12463 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12465 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
12466 Reformat alternatives and attributes so it is easier to identify
12467 which constraints/attributes go with which instruction.
12468 (mov<mode>_hardfloat32, FMOVE64): Likewise.
12469 (mov<mode>_softfloat32, FMOVE64): Likewise.
12470 (mov<mode>_hardfloat64, FMOVE64): Likewise.
12471 (mov<mode>_softfloat64, FMOVE64): Likewise.
12473 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
12475 * doc/extend.texi (PowerPC Built-in Functions): Rename this
12477 (Basic PowerPC Built-in Functions): The new name of the
12478 subsection previously known as "PowerPC Built-in Functions".
12479 (Basic PowerPC Built-in Functions Available on all Configurations):
12481 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
12482 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
12483 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
12484 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
12486 2018-05-11 Martin Jambor <mjambor@suse.cz>
12489 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
12492 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
12495 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
12497 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
12499 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
12500 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
12501 (ix86_handle_option): Handle -mwaitpkg.
12502 * config.gcc: New header.
12503 * config/i386/cpuid.h (bit_WAITPKG): New bit.
12504 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
12505 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
12507 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12508 OPTION_MASK_ISA_WAITPKG.
12509 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
12510 (ix86_option_override_internal): Add PTA_WAITPKG.
12511 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
12512 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
12513 IX86_BUILTIN_TPAUSE.
12514 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
12515 __builtin_ia32_umwait and __builtin_ia32_tpause.
12516 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
12517 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
12518 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
12519 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
12520 UNSPECV_TPAUSE): New.
12521 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
12522 * config/i386/i386.opt: Add -mwaitpkg.
12523 * config/i386/waitpkgintrin.h: New file.
12524 * config/i386/x86intrin.h: New header.
12525 * doc/invoke.texi: Add -mwaitpkg.
12527 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
12530 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
12532 (cortex-m0): Use armv6s-m isa.
12533 (cortex-m0plus): Likewise.
12534 (cortex-m1): Likewise.
12535 (cortex-m0.small-multiply): Likewise.
12536 (cortex-m0plus.small-multiply): Likewise.
12537 (cortex-m1.small-multiply): Likewise.
12539 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
12540 Jakub Jelinek <jakub@redhat.com>
12542 PR tree-optimization/85692
12543 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
12544 source permute as well.
12546 2018-05-11 Martin Liska <mliska@suse.cz>
12549 * doc/extend.texi: Document LLVM style format for no_sanitize
12552 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12554 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
12555 mode_supports_vsx_dform_quad to mode_supports_dq_form.
12556 (mode_supports_vsx_dform_quad): Likewise.
12557 (mode_supports_vmx_dform): Move these functions to be next to the
12558 other mode_supports functions.
12559 (mode_supports_dq_form): Likewise.
12560 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
12561 mode_supports_dq_form.
12562 (reg_offset_addressing_ok_p): Likewise.
12563 (offsettable_ok_by_alignment): Likewise.
12564 (rs6000_legitimate_offset_address_p): Likewise.
12565 (legitimate_lo_sum_address_p): Likewise.
12566 (rs6000_legitimize_address): Likewise.
12567 (rs6000_legitimize_reload_address): Likewise.
12568 (rs6000_secondary_reload_inner): Likewise.
12569 (rs6000_preferred_reload_class): Likewise.
12570 (rs6000_output_move_128bit): Likewise.
12572 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
12574 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
12575 Generate SImode target register for null target.
12576 <case IX86_BUILTIN_XGETBV>: Ditto.
12577 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
12578 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
12580 2018-05-10 Carl Love <cel@us.ibm.com>
12582 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
12583 dcbtt and dcbtstt if operands[2] is 0.
12585 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
12588 * config/i386/sse.md (usadv64qi): New expander.
12590 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
12592 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
12593 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
12594 -maltivec=be support.
12595 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
12596 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
12597 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
12598 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
12599 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
12600 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
12601 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
12602 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
12603 altivec_vsumsws): Adjust.
12604 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
12605 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
12606 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
12608 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
12609 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
12610 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
12611 (altivec_lve<VI_char>x): Delete expand.
12612 (*altivec_lve<VI_char>x_internal): Rename to...
12613 (altivec_lve<VI_char>x): ... this.
12614 (altivec_lvxl_<mode>): Delete expand.
12615 (*altivec_lvxl_<mode>_internal): Rename to ...
12616 (altivec_lvxl_<mode>): ... this.
12617 (altivec_stvxl_<mode>): Delete expand.
12618 (*altivec_stvxl_<mode>_internal): Rename to ...
12619 (altivec_stvxl_<mode>): ... this.
12620 (altivec_stve<VI_char>x): Delete expand.
12621 (*altivec_stve<VI_char>x_internal): Rename to ...
12622 (altivec_stve<VI_char>x): ... this.
12623 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
12624 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
12625 reduc_plus_scal_<mode>): Adjust.
12626 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
12628 (rs6000_cpu_cpp_builtins): Adjust.
12629 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
12630 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
12631 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
12632 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
12633 -maltivec=be support.
12634 (rs6000_split_vec_extract_var): Adjust.
12635 (rs6000_split_v4si_init): Adjust.
12636 (swap_selector_for_mode): Delete.
12637 (altivec_expand_lvx_be, altivec_expand_stvx_be,
12638 altivec_expand_stvex_be): Delete.
12639 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
12640 -maltivec=be support.
12641 (rs6000_gimple_fold_builtin): Ditto.
12642 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
12644 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
12645 (TARGET_DIRECT_MOVE_64BIT): Adjust.
12646 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
12647 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
12648 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
12649 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
12650 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
12651 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
12652 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
12653 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
12654 anonymous split): Adjust.
12655 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
12656 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
12658 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
12660 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
12661 when --with-gxx-include-dir is also specified.
12662 * configure: Regenerate.
12664 2018-05-09 Jim Wilson <jimw@sifive.com>
12667 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
12668 * config/riscv/t-withmultilib: New.
12669 * config/riscv/withmultilib.h: New.
12670 * doc/install.texi: Document RISC-V --with-multilib-list support.
12672 2018-05-09 Richard Biener <rguenther@suse.de>
12674 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
12676 (vect_bb_vectorization_profitable_p): Adjust. Compute
12677 actual scalar cost using the cost vector and the add_stmt_cost
12680 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12682 PR rtl-optimization/85645
12683 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
12684 in the REG_CFA_REGISTER note for LR, don't leave it empty.
12686 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12688 PR rtl-optimization/85645
12689 * shrink-wrap.c (spread_components): Return a boolean saying if
12690 anything was changed.
12691 (try_shrink_wrapping_separate): Iterate spread_components until
12692 nothing changes anymore.
12694 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12696 PR rtl-optimization/85645
12697 * regrename.c (build_def_use): Also kill the chains that include the
12698 destination of a REG_CFA_REGISTER note.
12700 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12702 PR rtl-optimization/85645
12703 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
12704 insn that has a REG_CFA_REGISTER note.
12706 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
12708 * cfgexpand.c (expand_clobber): New function.
12709 (expand_gimple_stmt_1): Use it.
12710 * tree-vect-stmts.c (vect_clobber_variable): New function,
12712 (vectorizable_simd_clone_call): ...here.
12713 (vectorizable_store): Emit a clobber either side of an
12714 IFN_STORE_LANES sequence.
12715 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
12717 2018-05-09 Tom de Vries <tom@codesourcery.com>
12720 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
12721 (define_insn "trap_if_false"): Add exit after trap.
12723 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
12725 PR rtl-optimization/85638
12726 * bb-reorder.c: Include common/common-target.h.
12727 (create_forwarder_block): New function extracted from...
12728 (fix_up_crossing_landing_pad): ...here. Rename into...
12729 (dw2_fix_up_crossing_landing_pad): ...this.
12730 (sjlj_fix_up_crossing_landing_pad): New function.
12731 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
12732 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
12733 from both partitions and exit the loop after one iteration.
12735 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
12738 * doc/extend.texi (PowerPC Built-in Functions): Rename this
12740 (Basic PowerPC Built-in Functions): The new name of the
12741 subsection previously known as "PowerPC Built-in Functions".
12742 (Basic PowerPC Built-in Functions Available on all Configurations):
12744 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
12746 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
12748 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
12750 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
12753 2018-05-08 Jim Wilson <jimw@sifive.com>
12755 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
12756 (LD_EMUL_SUFFIX): New.
12757 (LINK_SPEC): Use it.
12759 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
12761 * doc/extend.texi (PowerPC Built-in Functions): Rename this
12763 (Basic PowerPC Built-in Functions): The new name of the
12764 subsection previously known as "PowerPC Built-in Functions".
12765 (Basic PowerPC Built-in Functions Available on all Configurations):
12767 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
12769 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
12771 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
12773 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
12776 2018-05-08 Jakub Jelinek <jakub@redhat.com>
12779 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
12780 after cmpelim optimization.
12782 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
12784 * config.gcc: Support "goldmont".
12785 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
12786 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12787 PROCESSOR_GOLDMONT.
12788 * config/i386/i386.c (m_GOLDMONT): Define.
12789 (processor_target_table): Add "goldmont".
12790 (PTA_GOLDMONT): Define.
12791 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
12792 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
12793 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
12794 (fold_builtin_cpu): Add "goldmont".
12795 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
12796 (ix86_option_override_internal): Add "goldmont".
12797 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
12798 (processor_type): Add PROCESSOR_GOLDMONT.
12799 * config/i386/i386.md: Add CPU "glm".
12800 * config/i386/glm.md: New file.
12801 * config/i386/x86-tune.def: Add m_GOLDMONT.
12802 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
12804 2018-05-08 Jakub Jelinek <jakub@redhat.com>
12807 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
12809 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
12810 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
12811 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
12814 * config/i386/i386.c (ix86_fold_builtin): Handle
12815 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
12818 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
12819 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
12821 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
12824 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
12825 (check_arch): Likewise.
12826 (check_fpu): Return the result rather than printing it.
12827 (end arch): Fix operator precedence.
12828 (end cpu): Likewise.
12829 (END): Print the result from check_fpu.
12831 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
12832 Alan Hayward <alan.hayward@arm.com>
12833 David Sherwood <david.sherwood@arm.com>
12835 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
12836 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
12837 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
12838 (*fcmuo<mode>_and): New patterns.
12840 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
12842 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
12843 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
12844 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
12845 (cmp_op, sve_imm_con): New code attributes.
12846 (SVE_COND_INT_CMP, imm_con): Delete.
12847 (cmp_op): Remove above unspecs from int attribute.
12848 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
12850 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
12851 comparison-specific unspecs.
12852 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
12853 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
12854 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
12855 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
12856 (*vec_fcm<cmp_op><mode>): Rename to...
12857 (*fcm<cmp_op><mode>): ...this and adjust likewise.
12858 (*vec_fcmuo<mode>): Rename to...
12859 (*fcmuo<mode>): ...this and adjust likewise.
12860 (*pred_fcm<cmp_op><mode>): New pattern.
12861 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
12862 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
12864 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
12866 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
12867 (aarch64_emit_sve_predicated_cond): New function.
12868 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
12869 (aarch64_emit_unspec_cond_or): Replace with...
12870 (aarch64_emit_sve_or_conds): ...this new function. Use
12871 aarch64_emit_sve_ptrue_op for the individual comparisons and
12872 aarch64_emit_binop to OR them together.
12873 (aarch64_emit_inverted_unspec_cond): Replace with...
12874 (aarch64_emit_sve_inverted_cond): ...this new function. Use
12875 aarch64_emit_sve_ptrue_op for the comparison and
12876 aarch64_emit_unop to invert the result.
12877 (aarch64_expand_sve_vec_cmp_float): Update after the above
12878 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
12880 2018-05-07 Nathan Sidwell <nathan@acm.org>
12882 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
12883 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
12884 (Backwards Compatibility): Likewise.
12886 2018-05-07 Luis Machado <luis.machado@linaro.org>
12890 2018-05-07 Luis Machado <luis.machado@linaro.org>
12892 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12893 <prefetch_dynamic_strides>: New const bool field.
12894 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12895 prefetch_dynamic_strides.
12896 (exynosm1_prefetch_tune): Likewise.
12897 (thunderxt88_prefetch_tune): Likewise.
12898 (thunderx_prefetch_tune): Likewise.
12899 (thunderx2t99_prefetch_tune): Likewise.
12900 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
12902 (aarch64_override_options_internal): Update to set
12903 PARAM_PREFETCH_DYNAMIC_STRIDES.
12904 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12905 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12906 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12907 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12908 prefetch-dynamic-strides setting.
12910 2018-05-07 Luis Machado <luis.machado@linaro.org>
12912 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12913 <minimum_stride>: New const int field.
12914 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12915 minimum_stride field.
12916 (exynosm1_prefetch_tune): Likewise.
12917 (thunderxt88_prefetch_tune): Likewise.
12918 (thunderx_prefetch_tune): Likewise.
12919 (thunderx2t99_prefetch_tune): Likewise.
12920 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
12921 (aarch64_override_options_internal): Update to set
12922 PARAM_PREFETCH_MINIMUM_STRIDE.
12923 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12924 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12925 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12926 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12927 stride is constant and is below the minimum stride threshold.
12929 2018-05-07 Luis Machado <luis.machado@linaro.org>
12931 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
12934 2018-05-07 Luis Machado <luis.machado@linaro.org>
12936 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12937 <prefetch_dynamic_strides>: New const bool field.
12938 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12939 prefetch_dynamic_strides.
12940 (exynosm1_prefetch_tune): Likewise.
12941 (thunderxt88_prefetch_tune): Likewise.
12942 (thunderx_prefetch_tune): Likewise.
12943 (thunderx2t99_prefetch_tune): Likewise.
12944 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
12946 (aarch64_override_options_internal): Update to set
12947 PARAM_PREFETCH_DYNAMIC_STRIDES.
12948 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12949 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12950 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12951 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12952 prefetch-dynamic-strides setting.
12954 2018-05-07 Luis Machado <luis.machado@linaro.org>
12956 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12957 <minimum_stride>: New const int field.
12958 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12959 minimum_stride field.
12960 (exynosm1_prefetch_tune): Likewise.
12961 (thunderxt88_prefetch_tune): Likewise.
12962 (thunderx_prefetch_tune): Likewise.
12963 (thunderx2t99_prefetch_tune): Likewise.
12964 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
12965 (aarch64_override_options_internal): Update to set
12966 PARAM_PREFETCH_MINIMUM_STRIDE.
12967 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12968 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12969 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12970 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12971 stride is constant and is below the minimum stride threshold.
12973 2018-05-06 Jakub Jelinek <jakub@redhat.com>
12976 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
12977 the type is addressable. Don't force op into register if it has
12980 2018-05-05 Roland McGrath <mcgrathr@google.com>
12983 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
12984 any section for which we don't know a specific type it should have,
12985 regardless of name. Previously this was done only for the exact
12986 names ".init_array", ".fini_array", and ".preinit_array".
12987 (default_elf_asm_named_section): Add comment about
12988 relationship with default_section_type_flags and SECTION_NOTYPE.
12989 (get_section): Don't consider it a type conflict if one side has
12990 SECTION_NOTYPE and the other doesn't, as long as neither has the
12991 SECTION_BSS et al used in the default_section_type_flags logic.
12993 2018-05-05 Tom de Vries <tom@codesourcery.com>
12996 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
12997 (workaround_barsyncs): New function.
12998 (nvptx_reorg): Use workaround_barsyncs.
12999 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
13000 (define_expand "nvptx_membar_cta"): New define_expand.
13001 (define_insn "*nvptx_membar_cta"): New insn.
13003 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
13005 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
13006 To improve optimization opportunities.
13007 * builtin-types.def: The new needed builtin types for the above.
13009 2018-05-04 Richard Biener <rguenther@suse.de>
13011 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
13012 * gimple-ssa-store-merging.c
13013 (imm_store_chain_info::output_merged_store): Remove redundant create,
13014 release split_store vector contents on failure.
13015 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
13016 scalar stmt vector on cache hit.
13018 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
13020 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
13022 * config.gcc (powerpc-xilinx-eabi*): Remove.
13023 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
13025 (fusion_addis_mem_combo_load): Ditto.
13026 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
13028 (rs6000_cpu_cpp_builtins): Ditto.
13029 * config/rs6000/rs6000-linux.c
13030 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
13031 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
13032 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
13034 (rs6000_setup_reg_addr_masks): Ditto.
13035 (rs6000_init_hard_regno_mode_ok): Ditto.
13036 (rs6000_option_override_internal): Ditto.
13037 (legitimate_lo_sum_address_p): Ditto.
13038 (rs6000_legitimize_address): Ditto.
13039 (rs6000_legitimize_reload_address): Ditto.
13040 (rs6000_legitimate_address_p): Ditto.
13041 (abi_v4_pass_in_fpr): Ditto.
13042 (setup_incoming_varargs): Ditto.
13043 (rs6000_gimplify_va_arg): Ditto.
13044 (rs6000_split_multireg_move): Ditto.
13045 (rs6000_savres_strategy): Ditto.
13046 (rs6000_emit_prologue_components): Ditto.
13047 (rs6000_emit_epilogue_components): Ditto.
13048 (rs6000_emit_prologue): Ditto.
13049 (rs6000_emit_epilogue): Ditto.
13050 (rs6000_elf_file_end): Ditto.
13051 (rs6000_function_value): Ditto.
13052 (rs6000_libcall_value): Ditto.
13053 * config/rs6000/rs6000.h: Ditto.
13054 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
13055 (TARGET_MINMAX): ... this. New.
13056 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
13057 * config/rs6000/rs6000.md: Remove Xilinx FP support.
13058 (*movsi_internal1_single): Delete.
13059 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
13060 mfpu=, mxilinx-fpu): Delete.
13061 * config/rs6000/singlefp.h: Delete.
13062 * config/rs6000/sysv4.h: Remove Xilinx FP support.
13063 * config/rs6000/t-rs6000: Ditto.
13064 * config/rs6000/t-xilinx: Delete.
13065 * config/rs6000/titan.md: Adjust for fp_type removal.
13066 * config/rs6000/vsx.md: Remove Xilinx FP support.
13067 (VStype_simple): Delete.
13068 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
13069 * config/rs6000/xfpu.h: Delete.
13070 * config/rs6000/xfpu.md: Delete.
13071 * config/rs6000/xilinx.h: Delete.
13072 * config/rs6000/xilinx.opt: Delete.
13073 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
13074 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
13076 2018-05-04 Tom de Vries <tom@codesourcery.com>
13079 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
13082 2018-05-04 Richard Biener <rguenther@suse.de>
13084 PR middle-end/85627
13085 * tree-complex.c (update_complex_assignment): We are always in SSA form.
13086 (expand_complex_div_wide): Likewise.
13087 (expand_complex_operations_1): Likewise.
13088 (expand_complex_libcall): Preserve EH info of the original stmt.
13089 (tree_lower_complex): Handle removed blocks.
13090 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
13091 on complex multiplication and division libcall builtins.
13093 2018-05-04 Richard Biener <rguenther@suse.de>
13095 PR middle-end/85574
13096 * fold-const.c (negate_expr_p): Restrict negation of operand
13097 zero of a division to when we know that can happen without
13099 (fold_negate_expr_1): Likewise.
13101 2018-05-04 Jakub Jelinek <jakub@redhat.com>
13104 * real.h (real_nextafter): Declare.
13105 * real.c (real_nextafter): New function.
13106 * fold-const-call.c (fold_const_nextafter): New function.
13107 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
13108 CASE_CFN_NEXTTOWARD.
13109 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
13110 even when arg1_mode is different from arg0_mode.
13112 2018-05-03 Nathan Sidwell <nathan@acm.org>
13114 * doc/extend.texi (Deprecated Features): Remove
13115 -ffriend-injection.
13116 (Backwards Compatibility): Likewise.
13117 * doc/invoke.texi (C++ Language Options): Likewise.
13118 (C++ Dialect Options): Likewise.
13120 2018-05-03 Jakub Jelinek <jakub@redhat.com>
13123 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
13124 _mm512_mask_mullox_epi64): New intrinsics.
13126 2018-05-03 Tom de Vries <tom@codesourcery.com>
13129 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13130 dump files): Add offload-tree.
13132 2018-05-03 Richard Biener <rguenther@suse.de>
13134 PR tree-optimization/85615
13135 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
13136 to loops not nested in BBs loop father to avoid creating multi-entry
13139 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13141 PR tree-optimization/70291
13142 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
13143 arguments. Change return type to tree. Emit libcall as a new
13144 statement rather than replacing existing one when inplace_p is true.
13145 (expand_complex_multiplication_components): New function.
13146 (expand_complex_multiplication): Expand floating-point complex
13147 multiplication using the above.
13148 (expand_complex_division): Rename inner_type parameter to type.
13149 Update expand_complex_libcall call-site.
13150 (expand_complex_operations_1): Update expand_complex_multiplication
13151 and expand_complex_division call-sites.
13153 2018-05-02 Jakub Jelinek <jakub@redhat.com>
13156 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
13157 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
13158 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
13159 the highest significant bit of the shift count mask is clear. In
13160 check whether and[sq]i3 is needed verify that all significant bits
13161 of the shift count other than the highest are set.
13163 2018-05-02 Tom de Vries <tom@codesourcery.com>
13166 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
13167 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
13168 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
13169 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
13170 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
13171 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
13172 __builtin_goacc_parlevel_size.
13174 2018-05-02 Richard Biener <rguenther@suse.de>
13176 PR tree-optimization/85597
13177 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
13178 do not use split vect_get_vec_defs call but call vect_get_slp_defs
13181 2018-05-02 Tom de Vries <tom@codesourcery.com>
13184 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13185 dump files): Add ltrans-tree.
13187 2018-05-02 Tom de Vries <tom@codesourcery.com>
13190 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13191 dump files): Add wpa-ipa.
13193 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
13195 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
13196 powerpc*-*-linux*paired* target.
13197 * config/rs6000/750cl.h: Delete.
13198 * config/rs6000/paired.h: Delete.
13199 * config/rs6000/paired.md: Delete.
13200 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
13202 * config/rs6000/rs6000-builtin.def: Remove paired float support.
13203 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
13204 comment. Remove paired float support.
13205 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
13206 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
13208 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
13209 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
13211 * config/rs6000/rs6000.c: Remove paired float support.
13212 (paired_expand_vector_init, paired_expand_vector_move,
13213 paired_emit_vector_compare, paired_emit_vector_cond_expr,
13214 (paired_expand_lv_builtin, paired_expand_stv_builtin,
13215 paired_expand_builtin, paired_expand_predicate_builtin,
13216 paired_init_builtins): Delete.
13217 * config/rs6000/rs6000.h: Remove paired float support.
13218 * config/rs6000/rs6000.md: Remove paired float support.
13219 (move_from_CR_ov_bit): Delete.
13220 * config/rs6000/rs6000.opt (mpaired): Delete.
13221 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
13222 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
13224 2018-05-02 Richard Biener <rguenther@suse.de>
13226 PR middle-end/85567
13227 * gimplify.c (gimplify_save_expr): When in SSA form allow
13228 SAVE_EXPRs to compute to SSA vars.
13230 2018-05-02 Jakub Jelinek <jakub@redhat.com>
13233 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
13234 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
13235 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
13236 clobber operands[2], instead use a new pseudo. Formatting fixes.
13238 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
13240 PR tree-optimization/85586
13241 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
13242 exit early for statements in the same group if the accesses are
13245 2018-05-02 Tom de Vries <tom@codesourcery.com>
13248 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
13251 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
13253 PR tree-optimization/85143
13254 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
13256 2018-05-01 Tom de Vries <tom@codesourcery.com>
13259 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
13260 not found" error message.
13262 2018-05-01 Tom de Vries <tom@codesourcery.com>
13265 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
13266 * vec.c (test_ordered_remove_if): New function.
13267 (vec_c_tests): Call test_ordered_remove_if.
13268 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
13269 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
13270 * tree-vect-patterns.c (vect_pattern_recog_1): Use
13271 VEC_ORDERED_REMOVE_IF.
13273 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13275 PR tree-optimization/82665
13276 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
13277 pointer subtraction where arguments come from a memchr call.
13279 2018-05-01 Jakub Jelinek <jakub@redhat.com>
13281 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
13282 --push-state --as-needed and --pop-state instead of --as-needed and
13283 --no-as-needed if ld supports it.
13284 * configure: Regenerated.
13287 * doc/install.texi2html: Replace _002d with - and _002a with * in
13288 generated html files using sed.
13290 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13293 * gcc-rich-location.c (blank_line_before_p): New function.
13294 (use_new_line): New function.
13295 (gcc_rich_location::add_fixit_insert_formatted): New function.
13296 * gcc-rich-location.h
13297 (gcc_rich_location::add_fixit_insert_formatted): New function.
13299 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13301 * selftest.c (assert_streq): Rename "expected" and "actual" to
13302 "val1" and "val2". Extend NULL-handling to cover both inputs
13303 symmetrically, while still requiring both to be non-NULL for a pass.
13304 * selftest.h (assert_streq): Rename "expected" and "actual" to
13306 (ASSERT_EQ): Likewise.
13307 (ASSERT_EQ_AT): Likewise.
13308 (ASSERT_KNOWN_EQ): Likewise.
13309 (ASSERT_KNOWN_EQ_AT): Likewise.
13310 (ASSERT_NE): Likewise.
13311 (ASSERT_MAYBE_NE): Likewise.
13312 (ASSERT_MAYBE_NE_AT): Likewise.
13313 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
13314 the assertion to pass.
13315 (ASSERT_STREQ_AT): Likewise.
13317 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
13319 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
13320 interaction with -pie.
13322 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13324 * selftest.h: Fix alphabetization of per-source-file selftest
13327 2018-04-30 Jason Merrill <jason@redhat.com>
13329 PR c++/61982 - dead stores to destroyed objects.
13330 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
13333 2018-04-30 Jason Merrill <jason@redhat.com>
13335 * tree.c (build_clobber): New.
13336 * tree.h: Declare it.
13337 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
13339 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13341 * diagnostic-show-locus.c (layout::layout): Update for
13342 location_get_source_line returning a char_span.
13343 (struct char_span): Move to input.h.
13344 (struct correction): Update for fields in char_span becoming
13346 (struct source_line): Update for location_get_source_line
13347 returning a char_span.
13348 (layout::print_line): Likewise.
13349 * edit-context.c (edited_file::print_content): Likewise.
13350 (edited_file::print_diff_hunk): Likewise.
13351 (edited_file::print_run_of_changed_lines): Likewise.
13352 (edited_file::get_num_lines): Likewise.
13353 (edited_line::edited_line): Likewise.
13354 * final.c (asm_show_source): Likewise.
13355 * input.c (location_get_source_line): Convert return type
13356 from const char * to char_span, losing the final "line_len"
13358 (dump_location_info): Update for the above.
13359 (get_substring_ranges_for_loc): Likewise. Use a char_span
13360 when handling the literal within the line.
13361 (test_reading_source_line): Update for location_get_source_line
13362 returning a char_span.
13363 * input.h (class char_span): Move here from
13364 diagnostic-show-locus.c, converting from a struct to a class.
13365 Make data members private.
13366 (char_span::operator bool): New.
13367 (char_span::length): New.
13368 (char_span::get_buffer): New.
13369 (char_span::operator[]): New.
13370 (char_span::subspan): Make const.
13371 (char_span::xstrdup): New.
13372 (location_get_source_line): Convert return type from const char *
13373 to char_span, losing the final "line_size" param.
13375 2018-04-30 Jan Hubicka <jh@suse.cz>
13377 * lto-wrapper.c (ltrans_priorities): New static var.
13378 (cmp_priority): New.
13379 (run_gcc): Read priorities and if doing parallel build order
13380 the Makefile by them.
13382 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13384 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
13386 2018-04-30 Richard Biener <rguenther@suse.de>
13388 * tree-cfg.c (verify_address): Remove base argument, add
13389 flag whether to check TREE_ADDRESSABLE and do that.
13390 (verify_expr): Remove.
13391 (verify_types_in_gimple_reference): Add pieces from verify_expr.
13392 (verify_gimple_assign_single): Likewise.
13393 (verify_gimple_switch): Likewise.
13394 (verify_expr_location_1): Dereference tp once. Add (disabled)
13395 piece from verify_expr.
13396 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
13398 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
13400 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
13402 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
13404 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
13405 (small_data_pattern): Likewise.
13406 (arc_rewrite_small_data): Likewise.
13407 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
13408 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
13409 (get_symbol_alignment): New function.
13410 (legitimate_small_data_address_p): Likewise.
13411 (legitimate_scaled_address): Update, call
13412 legitimate_small_data_address_p.
13413 (output_sdata): New static variable.
13414 (arc_print_operand): Update how we handle small data operands.
13415 (arc_print_operand_address): Likewise.
13416 (arc_legitimate_address_p): Update, use
13417 legitimate_small_data_address_p.
13418 (arc_rewrite_small_data_p): Remove.
13419 (arc_rewrite_small_data_1): Likewise.
13420 (arc_rewrite_small_data): Likewise.
13421 (small_data_pattern): Likewise.
13422 (compact_sda_memory_operand): Update to use
13423 legitimate_small_data_address_p and get_symbol_alignment.
13424 (prepare_move_operands): Don't rewite sdata pattern.
13425 (prepare_extend_operands): Remove.
13426 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
13428 (zero_extendqisi2): Likewise.
13429 (zero_extendhisi2): Likewise.
13430 (extendqihi2): Likewise.
13431 (extendqisi2): Likewise.
13432 (extendhisi2): Likewise.
13433 (addsi3): Likewise.
13434 (subsi3): Likewise.
13435 (andsi3): Likewise.
13436 * config/arc/constraints.md (Usd): Change it to memory constraint.
13438 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
13440 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
13441 as source of std instructions.
13442 * config/arc/arc.md (movsi_insn): Update pattern predicate to
13443 allow 6-bit constants as source for store instructions.
13444 (movdi_insn): Update instruction pattern to allow 6-bit constants
13445 as source for store instructions.
13447 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
13449 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
13451 2018-04-30 Nathan Sidwell <nathan@acm.org>
13452 Sandra Loosemore <sandra@codesourcery.com>
13454 * dumpfile.c (dump_open): Allow '-' for stdout.
13455 * doc/invoke.texi (Developer Options): Document dump filename
13456 determination early. Document stdin/stdout selection.
13458 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
13460 Microblaze Target: PIC data text relative
13462 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
13463 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
13465 * config/microblaze/microblaze.h (microblaze_constant_address_p):
13466 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
13467 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
13468 New addressing mode for data-text relative position indepenedent code.
13469 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
13470 'ADDRESS_SYMBOLIC_TXT_REL'.
13471 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
13472 (microblaze_legitimate_pic_operand): Exclude function calls from
13473 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
13474 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
13476 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
13477 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
13478 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
13479 for 'address + offset'.
13480 (microblaze_expand_prologue): Add new function prologue call for
13482 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
13483 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
13484 table in case of TARGET_PIC_DATA_TEXT_REL.
13485 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
13486 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
13487 Add new macros 'UNSPEC_TEXT',
13488 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
13489 + exclude function calls from 'UNSPEC_PLT' in case of data text
13491 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
13492 new target hook for generating address diff vector tables in case of
13494 * doc/tm.texi : Regenerate.
13495 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
13496 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
13497 of addr diff vector generation.
13498 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
13499 target hook definition.
13500 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
13501 Add default function for generate_pic_addr_diff_vec -> flag_pic.
13502 * doc/invoke.texi (Add new pic option): Add new microblaze pic
13503 option for data text relative.
13505 2018-04-30 Richard Biener <rguenther@suse.de>
13507 * tree-chrec.h (evolution_function_is_constant_p): Remove
13509 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
13511 2018-04-30 Richard Biener <rguenther@suse.de>
13514 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
13516 2018-04-30 Richard Biener <rguenther@suse.de>
13518 PR tree-optimization/28364
13519 PR tree-optimization/85275
13520 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
13521 copying first exit test.
13523 2018-04-28 Mark Wielaard <mark@klomp.org>
13525 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
13526 dwarf_version >= 5.
13527 (dwarf_AT): Handle DW_AT_addr_base.
13528 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
13530 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
13533 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
13534 (*ashl<dwi>3_doubleword_mask_1): Ditto.
13535 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
13536 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
13538 2018-04-28 Richard Biener <rguenther@suse.de>
13540 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
13541 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
13542 to reflect use. Only add interesting stmts.
13544 2018-04-27 Martin Jambor <mjambor@suse.cz>
13547 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
13548 the jump function allows for passing through aggregate values.
13550 2018-04-27 David Malcolm <dmalcolm@redhat.com>
13552 * input.h (in_system_header_at): Convert from macro to inline
13554 (from_macro_expansion_at): Likewise.
13555 (from_macro_definition_at): Likewise.
13557 2018-04-27 Jeff Law <law@redhat.com>
13559 * config.gcc: Mark tile* targets as deprecated/obsolete.
13561 2018-04-27 Richard Biener <rguenther@suse.de>
13563 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
13566 2018-04-27 Richard Biener <rguenther@suse.de>
13568 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
13570 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
13572 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
13573 with Yd constraint. Set "preferred_for_speed" attribute from
13574 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
13575 with Yd constraint.
13576 (*movdi_internal): Ditto.
13577 (movti_interunit splitters): Remove
13578 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
13579 (movdi_interunit splitters): Ditto.
13580 * config/i386/constraints.md (Ye): Remove.
13581 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
13583 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13586 * config/aarch64/constraints.md (Usg): Limit to 31.
13587 (Usj): Limit to 63.
13589 2018-04-27 Jakub Jelinek <jakub@redhat.com>
13591 PR tree-optimization/85529
13592 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
13593 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
13594 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
13595 zero extension or masking of the MSB bit.
13596 (optimize_range_tests): Add FIRST_BB argument, pass it through
13597 to optimize_range_tests_var_bound.
13598 (maybe_optimize_range_tests, reassociate_bb): Adjust
13599 optimize_range_tests callers.
13601 2018-04-26 Richard Biener <rguenther@suse.de>
13602 Jakub Jelinek <jakub@redhat.com>
13604 * cgraph.h (symbol_table): Just declare debug method here.
13605 * symtab.c (symbol_table::debug): Define.
13607 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
13609 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
13611 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
13613 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
13614 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
13615 (*movdi_internal): Substitute Yi and Yj constraint with x
13616 and Ym and Yn constraint with y constraint. Update "isa"
13617 attribute and set "preferred_for_speed" attribute from
13618 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
13619 (*movsi_internal): Ditto.
13620 (*movdf_internal): Ditto.
13621 (*movsf_internal): Ditto.
13622 (*zero_extendsidi2): Ditto.
13623 * config/i386/sse.md (vec_set<mode>_0): Ditto.
13624 (sse2_loadld): Ditto.
13625 (*vec_extract<ssevecmodelower>_0): Ditto.
13626 (*vec_extractv4si_0_zext_sse4): Ditto.
13627 (vec_concatv2di): Ditto.
13628 (*vec_dup<mode>): Ditto.
13629 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
13630 * config/i386/constraints.md (Yi): Remove.
13635 2018-04-26 Nathan Sidwell <nathan@acm.org>
13637 * dumpfile.c (dump_open): New.
13638 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
13639 (dump_finish): Detect stdio/stderr by value not name.
13641 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
13643 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
13645 2018-04-26 Tom de Vries <tom@codesourcery.com>
13648 * config/nvptx/nvptx.c (verify_neutering_jumps)
13649 (verify_neutering_labels): New function
13650 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
13652 2018-04-26 Tom de Vries <tom@codesourcery.com>
13655 * config/nvptx/nvptx.c (needs_neutering_p): New function.
13656 (nvptx_single): Use needs_neutering_p to skip over insns that do not
13659 2018-04-26 Richard Biener <rguenther@suse.de>
13660 Tom de Vries <tom@codesourcery.com>
13663 * lto-streamer-out.c (output_function): Fixup loops if required to match
13664 discovery done in the reader.
13666 2018-04-26 Richard Biener <rguenther@suse.de>
13668 PR tree-optimization/85116
13669 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
13670 have a loop exit from the single latch predecessor. Remove
13671 case of header with just condition.
13672 (ch_base::copy_headers): Exclude infinite loops from any
13674 (pass_ch::execute): Record exits.
13676 2018-04-26 Richard Biener <rguenther@suse.de>
13678 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
13679 prologue cost vector and pass it to vect_get_load_cost.
13680 (vect_get_peeling_costs_all_drs): Likewise.
13681 (vect_peeling_hash_get_lowest_cost): Likewise.
13682 (vect_enhance_data_refs_alignment): Likewise.
13684 2018-04-26 Richard Biener <rguenther@suse.de>
13686 PR middle-end/85450
13687 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
13688 checking of integer<->pointer conversions.
13689 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
13690 sign-/zero-extending pointer types.
13691 (expand_omp_for_static_chunk): Likewise.
13693 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
13694 Jean Lee <xiaoyur347@gmail.com>
13696 * config/mips/mips.c (mips_asan_shadow_offset): New function.
13697 (TARGET_ASAN_SHADOW_OFFSET): Define.
13698 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
13699 true for -fsanitize=address.
13701 2018-04-25 Mark Wielaard <mark@klomp.org>
13703 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
13706 2018-04-25 Jakub Jelinek <jakub@redhat.com>
13708 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
13709 than "alu", remove explicit "memory" and "imm_disp" attributes.
13710 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
13712 PR middle-end/85414
13713 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
13714 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
13715 gen_lowpart_no_emit.
13717 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
13720 * config/i386/i386.c (ix86_expand_builtin): Change memory
13721 operand to XI, extend p0 to Pmode.
13722 * config/i386/i386.md: Change unspec volatile and operand
13723 1 mode to XI, change operand 0 mode to P.
13725 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13727 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
13728 GET_MODE_MASK before any checking.
13729 (nds32_can_use_bset_p): Likewise.
13730 (nds32_can_use_btgl_p): Likewise.
13732 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13734 * config/nds32/nds32-doubleword.md: New define_split pattern for
13735 illegal register number.
13737 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13739 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
13741 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13743 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
13745 2018-04-25 Richard Biener <rguenther@suse.de>
13747 * lto-streamer.h (LTO_major_version): Bump to 8.
13749 2018-04-25 Jakub Jelinek <jakub@redhat.com>
13751 * BASE-VER: Set to 9.0.0.
13753 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
13755 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
13756 in __abskf2 and __powikf2.
13758 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13761 * config/aarch64/constraints.md (Usg, Usj): New constraints.
13762 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
13763 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
13764 Use the above on operand 2. Reindent.
13765 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
13767 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
13770 * common/config/i386/i386-common.c (ix86_handle_option): Don't
13772 * config/i386/i386.opt (mcet): Removed.
13773 * doc/install.texi: Remove -mcet documentation.
13774 * doc/invoke.texi: Likewise.
13776 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
13779 * doc/install.texi: Remove -mcet from bootstrap-cet.
13781 2018-04-24 Jakub Jelinek <jakub@redhat.com>
13784 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
13785 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
13789 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
13790 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
13791 containing a CONST_VECTOR.
13793 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
13795 * doc/install.texi: Update newlib dependency for nvptx.
13797 2018-04-24 Jakub Jelinek <jakub@redhat.com>
13800 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
13801 instead of INTVAL when shifting x left.
13803 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
13805 PR tree-optimization/85478
13806 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
13807 vect_grouped_store_supported for single element vectors.
13809 2018-04-24 Richard Biener <rguenther@suse.de>
13812 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
13813 load cost increase to the case of non-constant step.
13815 2018-04-24 Jakub Jelinek <jakub@redhat.com>
13818 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
13819 destination if any_malformed_asm.
13821 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
13823 PR middle-end/85496
13824 * expr.c (store_field): In the bitfield case, if the value comes from
13825 a function call and is returned in registers by means of a PARALLEL,
13826 do not change the mode of the temporary unless BLKmode and VOIDmode.
13828 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
13830 PR rtl-optimization/85423
13831 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
13832 dependencies to debug insns when the previous insn is non-debug.
13834 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
13836 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
13837 enums into a single definition.
13838 (fls): Fix predicates and printing.
13841 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
13843 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
13844 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
13845 and short u6 immediate.
13846 (check_if_valid_sleep_operand): Remove.
13847 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
13849 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13851 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
13852 flag_always_save_lp condition.
13853 * config/nds32/nds32.opt (malways-save-lp): New option.
13855 2018-04-22 Shiva Chen <shiva0217@gmail.com>
13857 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
13858 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
13859 * config/nds32/nds32.h
13860 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
13861 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
13863 2018-04-22 Shiva Chen <shiva0217@gmail.com>
13865 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
13866 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
13868 2018-04-22 Shiva Chen <shiva0217@gmail.com>
13869 Chung-Ju Wu <jasonwucj@gmail.com>
13871 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
13873 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
13874 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
13876 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13878 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
13880 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13882 * config/nds32/nds32-protos.h (nds32_data_alignment,
13883 nds32_local_alignment): Declare.
13884 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
13885 nds32_local_alignment): New functions.
13886 (TARGET_CONSTANT_ALIGNMENT): Define.
13887 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
13889 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13891 * config/nds32/nds32.c
13892 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
13893 (TARGET_MODES_TIEABLE_P): Likewise.
13895 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13897 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
13898 level Ofast and Og.
13900 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
13901 Chung-Ju Wu <jasonwucj@gmail.com>
13903 * config/nds32/constants.md (unspec_volatile_element): Add enum values
13904 for unaligned access.
13905 * config/nds32/nds32-intrinsic.c: Implementation of expanding
13907 * config/nds32/nds32-intrinsic.md: Likewise.
13908 * config/nds32/nds32_intrinsic.h: Likewise.
13909 * config/nds32/nds32.h (nds32_builtins): Likewise.
13910 * config/nds32/nds32.opt (munaligned-access): New option.
13911 * config/nds32/nds32.c (nds32_asm_file_start): Display
13912 flag_unaligned_access status.
13914 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
13916 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
13917 -mno-relax is present.
13918 * config/riscv/linux.h (LINK_SPEC): Ditto.
13920 2018-04-20 Martin Sebor <msebor@redhat.com>
13923 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
13925 (gimple_fold_builtin_stxcpy_chk): Same.
13926 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
13928 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
13931 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
13932 __powikf2 when long double is IEEE 128-bit.
13934 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
13936 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
13937 step to make sure stack always aligned.
13939 2018-04-20 Carl Love <cel@us.ibm.com>
13942 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
13943 size check for arg0.
13945 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
13946 Tom de Vries <tom@codesourcery.com>
13949 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
13950 Emit insns for calls too.
13951 (nvptx_find_par): Always look for worker-level predecessor insn.
13952 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
13954 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
13955 (nvptx_process_pars): Propagate frames for calls.
13957 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
13960 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
13962 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
13963 (ix86_handle_option): Don't handle OPT_mibt.
13964 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
13966 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
13968 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
13969 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
13970 (ix86_target_macros): Define __CET__ with flag_cf_protection
13971 for -fcf-protection.
13972 * config/i386/i386.c (isa2_opts): Remove -mibt.
13973 * config/i386/i386.h (TARGET_IBT): Removed.
13974 (TARGET_IBT_P): Likewise.
13975 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
13976 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
13977 * config/i386/i386.opt (mcet): Update help message.
13978 (mshstk): Likewise.
13980 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
13981 -mcet as an alias for -mshstk.
13983 2018-04-20 Richard Biener <rguenther@suse.de>
13985 PR middle-end/85475
13986 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
13987 complexity by forcing a single use of the multiply operand.
13989 2018-04-20 Martin Jambor <mjambor@suse.cz>
13992 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
13993 recursion dependency to only apply to non-clones.
13995 2018-04-20 Martin Jambor <mjambor@suse.cz>
13998 * ipa-cp.c (create_specialized_node): Check that clones of
13999 self-recursive edges exist during IPA-CP.
14001 2018-04-19 Toon Moene <toon@moene.org>
14003 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
14006 2018-04-19 Jakub Jelinek <jakub@redhat.com>
14008 PR tree-optimization/85467
14009 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
14010 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
14011 VECTOR_CST element to type.
14013 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
14016 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
14017 * config/i386/i386.md (builtin_setjmp_setup): Removed.
14018 (builtin_longjmp): Likewise.
14019 (save_stack_nonlocal): New pattern.
14020 (restore_stack_nonlocal): Likewise.
14022 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
14025 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
14026 Replace ASM_OUTPUT_LABEL with fprintf.
14028 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
14031 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
14032 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
14033 * config/i386/i386-c.c (ix86_target_macros_internal): Also
14034 define __IBT__ and __SHSTK__ for -fcf-protection.
14035 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
14037 (ix86_trampoline_init): Likewise.
14038 (x86_output_mi_thunk): Likewise.
14039 (ix86_notrack_prefixed_insn_p): Likewise.
14040 (ix86_option_override_internal): Don't disallow -fcf-protection.
14041 * config/i386/i386.md (rdssp<mode>): Also enable for
14043 (incssp<mode>): Likewise.
14044 (nop_endbr): Likewise.
14045 * config/i386/i386.opt (mcet): Change help message to built-in
14048 (mshstk): Likewise.
14049 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
14050 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
14051 enable CET built-in functions.
14053 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
14055 * common/config/i386/i386-common.c
14056 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
14057 OPTION_MASK_ISA_MOVDIRI_UNSET,
14058 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
14059 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
14060 * config.gcc (movdirintrin.h): New header.
14061 * config/i386/cpuid.h (bit_MOVDIRI,
14062 bit_MOVDIR64B): New bits.
14063 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
14065 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
14066 (VOID, PVOID, PCVOID)): New function types.
14067 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
14068 __builtin_ia32_directstoreu_u64,
14069 __builtin_ia32_movdir64b): New builtins.
14070 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
14071 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
14073 (ix86_valid_target_attribute_inner_p): Ditto.
14074 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
14075 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
14076 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
14077 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
14078 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
14079 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
14080 (movdiri<mode>, movdir64b_<mode>): New.
14081 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
14082 * config/i386/immintrin.h: Include movdirintrin.h.
14083 * config/i386/movdirintrin.h: New file.
14084 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
14086 2018-04-19 Richard Biener <rguenther@suse.de>
14088 PR middle-end/85455
14089 * cfg.c (clear_bb_flags): When loop state says we have
14090 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
14092 2018-04-19 Richard Biener <rguenther@suse.de>
14094 PR tree-optimization/84737
14095 * tree-vect-data-refs.c (vect_copy_ref_info): New function
14096 copying restrict info.
14097 (vect_setup_realignment): Use it.
14098 * tree-vectorizer.h (vect_copy_ref_info): Declare.
14099 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
14100 the first DR to all generated stores.
14101 (vectorizable_load): Likewise for loads.
14103 2018-04-19 Jakub Jelinek <jakub@redhat.com>
14105 PR tree-optimization/85446
14106 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
14107 the integral and pointer types to have the same precision.
14109 * doc/install.texi: Document --disable-cet being the default and
14112 2018-04-18 Martin Liska <mliska@suse.cz>
14114 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
14117 2018-04-18 Martin Liska <mliska@suse.cz>
14120 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
14123 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
14124 arguments if they are comparable.
14126 2018-04-18 Martin Liska <mliska@suse.cz>
14129 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
14132 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
14135 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
14138 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
14139 ENDBR after calling __morestack.
14141 2018-04-18 David Malcolm <dmalcolm@redhat.com>
14144 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
14145 by using gcc_base_ver to generate a gcc_driver_version, and use
14146 it when generating GCC_DRIVER_NAME.
14147 * configure: Regenerate.
14149 2018-04-18 Jakub Jelinek <jakub@redhat.com>
14152 * config.gcc: Obsolete powerpc*-*-*spe*.
14154 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14157 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
14158 (stabstr_D): Change type of unum from unsigned int to
14159 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
14162 2018-04-17 Jim Wilson <jimw@sifive.com>
14165 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
14166 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
14167 Set arg_pointer_offset after using pretend_args_size.
14169 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14171 PR rtl-optimization/85431
14172 * dse.c (record_store): Ignore zero width stores.
14175 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
14176 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
14177 __builtin_stack_restore rather than after it.
14178 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
14179 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
14180 argument instead of virtual_dynamic_stack_rtx.
14182 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
14184 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
14186 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14187 Add note to error message to explain internal mapping of overloaded
14188 built-in function name to non-overloaded built-in function name.
14189 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
14192 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14195 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
14196 where the inputs overlap with the output.
14198 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14200 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
14201 (=v, v) alternative and explicit "memory" attribute.
14202 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
14203 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14205 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
14206 "sselog1" type instead of "sselog".
14207 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
14208 "sselog". Remove explicit "memory" attribute.
14209 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
14210 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14212 (vec_extract_hi_v32hi): Merge all alternatives into one, use
14213 "sselog1" type instead of "sselog". Remove explicit "memory"
14215 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
14216 use "sselog1" type instead of "sselog". Remove explicit "memory"
14218 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
14219 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14221 (vec_extract_hi_v64qi): Merge all alternatives into one, use
14222 "sselog1" type instead of "sselog". Remove explicit "memory"
14224 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
14225 use "sselog1" type instead of "sselog". Remove explicit "memory"
14229 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
14231 PR middle-end/85414
14232 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
14235 2018-04-17 Martin Jambor <mjambor@suse.cz>
14238 * ipa-cp.c (create_specialized_node): Call
14239 expand_all_artificial_thunks if necessary.
14241 2018-04-17 Martin Liska <mliska@suse.cz>
14244 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
14245 in message, remote space in between '_G' and '('.
14247 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14250 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
14251 avx512f_vmcmp<mode>3<round_saeonly_name>,
14252 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
14253 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
14254 avx512f_rndscale<mode><round_saeonly_name>,
14255 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
14256 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
14257 Use %<iptr>2 instead of %2 for -masm=intel.
14258 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
14259 avx512f_vcvttss2usi<round_saeonly_name>,
14260 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
14262 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
14263 avx512f_vcvttsd2usi<round_saeonly_name>,
14264 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
14265 Use %q1 instead of %1 for -masm=intel.
14266 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
14267 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
14268 of %3 for -masm=intel.
14269 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
14271 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
14273 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
14275 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
14276 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
14278 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
14280 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
14281 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
14282 %g1 and one with %0 and %1.
14283 (avx512er_vmrcp28<mode><round_saeonly_name>,
14284 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
14285 %1 for -masm=intel.
14286 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
14287 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
14288 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
14289 of %0 and %{%4%} for -masm=intel.
14290 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
14291 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
14292 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
14293 order of %0 and %{%5%}%{z%} for -masm=intel.
14295 2018-04-17 Jan Hubicka <jh@suse.cz>
14298 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
14300 2018-04-17 Martin Liska <mliska@suse.cz>
14303 * multiple_target.c (create_dispatcher_calls): Set apostrophes
14304 for target_clone error message. Make default implementation
14305 clone to be a local declaration.
14306 (separate_attrs): Add new argument and check for an empty
14308 (expand_target_clones): Handle it.
14309 (ipa_target_clone): Make redirection just for target_clones
14312 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
14313 Tom de Vries <tom@codesourcery.com>
14315 PR middle-end/84955
14316 * omp-expand.c (expand_oacc_for): Add dummy false branch for
14317 tiled basic blocks without omp continue statements.
14319 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
14322 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
14323 vec_extract expression as having side effects to make sure it gets
14326 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
14329 * config/i386/i386.c (get_builtin_code_for_version): Check
14332 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
14335 * config.gcc: Support "skylake".
14336 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14338 * config/i386/i386.c (m_SKYLAKE): Define.
14339 (processor_target_table): Add "skylake".
14340 (ix86_option_override_internal): Add "skylake".
14341 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
14342 PROCESSOR_CANNONLAKE.
14343 (get_builtin_code_for_version): Fix priority for
14344 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
14345 PROCESSOR_SKYLAKE-AVX512.
14346 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
14347 (processor_type): Add PROCESSOR_SKYLAKE.
14349 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
14350 Jason Merrill <jason@redhat.com>
14353 * convert.c (convert_to_integer_1): Use direct recursion for
14354 enumeral types and types with a precision less than the number
14355 of bits in their mode.
14357 2018-04-16 Julia Koval <julia.koval@intel.com>
14360 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
14361 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
14363 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
14366 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
14367 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
14368 and -mno-direct-move.
14370 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
14373 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
14374 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
14375 Ensure negative shifts result in {0}.
14377 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
14379 PR rtl-optimization/79916
14380 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
14381 regs (if any) to define how to gnerate SD moves when LRA is in
14384 2018-04-13 Jakub Jelinek <jakub@redhat.com>
14386 PR rtl-optimization/85393
14387 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
14388 * except.c (expand_dw2_landing_pad_for_region): Make static.
14389 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
14390 a label and unconditional jump to old_bb, rather than
14391 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
14394 PR rtl-optimization/85376
14395 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
14396 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
14397 instead of a specific value.
14399 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
14400 Bin Cheng <bin.cheng@arm.com>
14402 PR tree-optimization/82965
14403 PR tree-optimization/83991
14404 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
14405 by_profile_only parameter.
14406 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
14407 information if the loop was predicted to iterate too many times.
14408 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
14410 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
14413 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
14416 2018-04-13 Martin Liska <mliska@suse.cz>
14417 Jakub Jelinek <jakub@redhat.com>
14419 PR middle-end/81657
14420 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
14421 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
14422 * builtins.c (expand_builtin_memory_copy_args): Use
14423 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
14424 handle dest_addr == pc_rtx.
14426 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
14429 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
14430 asked to not generate direct moves.
14431 (fix_trunc<mode>si2_stfiwx): Similar.
14432 (fix_trunc<mode>si2_internal): Similar.
14434 2018-04-12 Jakub Jelinek <jakub@redhat.com>
14437 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
14438 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
14439 lookup if dest in some wider mode is known to be const0_rtx and
14440 if so, record permanent equivalence for it to be ZERO_EXTEND of
14441 the narrower mode destination.
14443 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
14445 * lto-streamer-out.c (output_function): Revert 259346.
14446 * omp-expand.c (expand_oacc_for): Likewise.
14448 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
14450 PR rtl-optimization/85354
14451 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
14452 * sel-sched.c (sel_global_init): ... here.
14454 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
14457 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
14458 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
14459 mode for PE-COFF targets.
14460 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
14461 (i386_pe_asm_lto_end): Likewise.
14462 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
14463 (TARGET_ASM_LTO_END): Likewise.
14464 * config/i386/winnt.c (saved_debug_info_level): New static variable.
14465 (i386_pe_asm_lto_start): New function.
14466 (i386_pe_asm_lto_end): Likewise.
14468 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
14469 Richard Biener <rguenther@suse.de>
14471 PR middle-end/84955
14472 * lto-streamer-out.c (output_function): Fix CFG loop state before
14474 * omp-expand.c (expand_oacc_for): Handle calls to internal
14475 functions like regular functions.
14477 2018-04-12 Richard Biener <rguenther@suse.de>
14480 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
14481 for the early LTO debug to properly generate references to it
14482 during DIE emission. Do not re-use that for the skeleton for
14484 (dwarf2out_early_finish): Likewise.
14486 2018-04-12 Jakub Jelinek <jakub@redhat.com>
14489 * config/i386/sse.md
14490 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
14491 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
14492 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
14493 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
14494 and output is a reg, avoid creating invalid lowpart subreg, but
14495 instead split into a 512-bit move. Don't split if not AVX512VL,
14496 input is xmm16+ reg and output is a mem.
14497 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
14498 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
14499 xmm16+ reg and output is a mem.
14501 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14503 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
14504 also for flag_dwarf2_cfi_asm.
14506 2018-04-12 Jakub Jelinek <jakub@redhat.com>
14508 PR rtl-optimization/85342
14509 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
14510 a bool scalar var inside of the loop instead. Don't try to update
14511 recog_data.operand after failed apply_change_group.
14513 2018-04-12 Tom de Vries <tom@codesourcery.com>
14516 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
14517 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
14518 array with flexible array member as array without given dimension.
14519 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
14520 argument for undefined param to true.
14522 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
14525 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
14526 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
14527 from PowerPC section.
14528 * config/rs6000/sysv4.opt (mcall-): Improve help text.
14529 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
14530 help text that is too long.
14531 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
14532 help text that is too long.
14533 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
14534 help text that is too long.
14536 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
14538 * config/alpha/alpha.md (stack_probe_internal): Rename
14539 from "probe_stack". Update all callers.
14541 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
14543 PR rtl-optimization/84566
14544 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
14545 sched_macro_fuse_insns.
14547 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
14550 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
14551 (compute_block_dependences): ... from here.
14553 2018-04-11 Jakub Jelinek <jakub@redhat.com>
14555 PR tree-optimization/85331
14556 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
14557 from int to HOST_WIDE_INT.
14559 2018-04-11 Martin Jambor <mjambor@suse.cz>
14562 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
14563 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
14564 not the same as the source val.
14565 (cgraph_edge_brings_value_p): New parameter.
14566 (gather_edges_for_value): Pass destination value to
14567 cgraph_edge_brings_value_p.
14568 (perhaps_add_new_callers): Likewise.
14569 (get_info_about_necessary_edges): Likewise and exclude values brought
14570 only by self-recursive edges.
14571 (create_specialized_node): Redirect only clones of self-calling edges.
14572 (+self_recursive_pass_through_p): New function.
14573 (find_more_scalar_values_for_callers_subset): Use it.
14574 (find_aggregate_values_for_callers_subset): Likewise.
14575 (known_aggs_to_agg_replacement_list): Removed.
14576 (decide_whether_version_node): Re-calculate known constants for all
14577 remaining context clones.
14579 2018-04-11 Richard Biener <rguenther@suse.de>
14582 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
14583 from early DWARF output.
14584 (dwarf2out_early_finish): Output line info unconditionally into
14585 early DWARF and add reference to it.
14587 2018-04-11 Jakub Jelinek <jakub@redhat.com>
14590 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
14591 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
14592 other than V2DFmode using iptr mode attribute.
14593 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
14595 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
14597 PR rtl-optimization/84659
14598 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
14600 2018-04-11 Jakub Jelinek <jakub@redhat.com>
14603 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
14605 (output_loc_list): Pass address of a dummy size variable even in the
14606 locview handling loop.
14607 (index_location_lists): Add comment on why skip_loc_list_entry can't
14610 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
14613 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
14616 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
14619 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
14620 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
14621 and -mstring-compare-inline-limit.
14623 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
14626 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
14627 for stack clash protection in a register whenever we need it to be in
14630 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
14632 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
14633 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
14635 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
14638 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
14640 (mlong-double-): Ditto.
14641 * config/rs6000/sysv4.opt (msdata=): Ditto.
14642 (mtls-size=): Ditto.
14644 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
14646 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
14647 erroneous entries for
14648 "vector int vec_ldl (int, long int *)", and
14649 "vector unsigned int vec_ldl (int, unsigned long int *)".
14650 Add comments and entries for
14651 "vector bool char vec_ldl (int, bool char *)",
14652 "vector bool short vec_ldl (int, bool short *)",
14653 "vector bool int vec_ldl (int, bool int *)",
14654 "vector bool long long vec_ldl (int, bool long long *)",
14655 "vector pixel vec_ldl (int, pixel *)",
14656 "vector long long vec_ldl (int, long long *)",
14657 "vector unsigned long long vec_ldl (int, unsigned long long *)".
14658 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
14659 type tree bool_long_long_type_node and correct definition of
14660 bool_V2DI_type_node to make reference to this new type tree.
14661 (rs6000_mangle_type): Replace erroneous reference to
14662 bool_long_type_node with bool_long_long_type_node.
14663 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
14664 comments to emphasize sign distinctions for char and int types and
14665 replace RS6000_BTI_bool_long constant with
14666 RS6000_BTI_bool_long_long constant. Also add comment to restrict
14667 use of RS6000_BTI_pixel.
14668 (bool_long_type_node): Remove this macro definition.
14669 (bool_long_long_type_node): New macro definition
14671 2018-04-10 Jakub Jelinek <jakub@redhat.com>
14673 PR rtl-optimization/85300
14674 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
14675 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
14676 simplify_unary_operation fails.
14678 2018-04-10 Martin Liska <mliska@suse.cz>
14680 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
14681 cgraph_edge and ipa_ref.
14683 2018-04-10 Jakub Jelinek <jakub@redhat.com>
14687 * config/i386/sse.md
14688 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
14689 computation of the VEC_MERGE selector from mask.
14690 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
14691 Fix decoding of the VEC_MERGE selector into mask.
14693 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
14695 PR tree-optimization/85286
14696 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
14698 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
14700 * final.c (final_1): Set insn_last_address as well as
14701 insn_current_address.
14703 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14706 * explow.c (emit_stack_probe): Call validize_mem on memory location
14707 before passing it to gen_probe_stack. Create address operand and
14708 legitimize it for the probe_stack_address case.
14710 2018-04-09 Jan Hubicka <jh@suse.cz>
14713 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
14714 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
14715 * tree.c (free_lang_data_in_type): Fix handling of binfos;
14717 (free_lang_data): Rebuild type inheritance graph.
14719 2018-04-09 Martin Sebor <msebor@redhat.com>
14721 * invoke.texi (-finline-small-functions): Mention other optimization
14723 (-findirect-inlining, -fpartial-inlining): Same.
14724 (-finline-functions-called-once): Same.
14725 (-freorder-blocks-and-partition): Same.
14727 2018-04-09 Jan Hubicka <jh@suse.cz>
14730 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
14731 jumps; choose last target that matches the criteria (i.e.
14732 no partition changes for non-crossing jumps).
14733 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
14734 support for redirecting crossing jumps to non-crossing.
14736 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
14738 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
14739 also for naked functions.
14741 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
14743 * config/arc/arc.md (add_shift): New pattern.
14744 (add_shift2): Likewise.
14745 (sub_shift): Likewise.
14746 (sub_shift_cmp0_noout): Likewise.
14747 (compare_si_ashiftsi): Likewise.
14748 (xbfu_cmp0_noout): New combine pattern.
14749 (xbfu_cmp0"): Likewise.
14750 (movsi_set_cc_insn): Place the predicable variant first.
14751 (commutative_binary_cmp0_noout): Remove clobber.
14752 (commutative_binary_cmp0): New pattern.
14753 (noncommutative_binary_cmp0): Likewise.
14754 (noncommutative_binary_cmp0_noout): Likewise.
14755 (noncommutative_binary_comparison_result_used): Removed.
14756 (rsub_cmp0): New pattern.
14757 (rsub_cmp0_noout): Likewise.
14758 (extzvsi): Changed, keep only meaningful variants.
14759 (SQH, SEZ): New iterators.
14760 (SQH_postfix): New mode attribute.
14761 (SEZ_prefix): New code attribute.
14762 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
14763 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
14764 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
14765 of numerical value.
14766 (noncommutative_operator): Check the availability of barrel
14769 2018-04-09 Richard Biener <rguenther@suse.de>
14771 PR tree-optimization/85284
14772 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
14773 Only use the niter constraining form of simple_iv when the exit
14774 is always executed.
14776 2018-04-09 Tom de Vries <tom@codesourcery.com>
14779 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
14780 (define_expand "*memory_barrier"): New define_expand.
14781 (define_insn "memory_barrier"): New insn.
14783 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14785 PR rtl-optimization/80463
14786 PR rtl-optimization/83972
14787 PR rtl-optimization/83480
14789 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
14790 correct producer for the insn.
14791 (tidy_control_flow): Fixup seqnos in case of debug insns.
14793 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14795 PR rtl-optimization/83913
14797 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
14798 different sched-times when merging exprs.
14800 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14802 PR rtl-optimization/83962
14804 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
14805 tidy_fallthru_edge and tidy_control_flow.
14807 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14809 PR rtl-optimization/83530
14811 * sel-sched.c (force_next_insn): New global variable.
14812 (remove_insn_for_debug): When force_next_insn is true, also leave only
14813 next insn in the ready list.
14814 (sel_sched_region): When the region wasn't scheduled, make another pass
14815 over it with force_next_insn set to 1.
14817 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
14819 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
14821 * config/nds32/constants.md (unspec_volatile_element): Add enum values
14822 for interrupt control.
14823 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
14824 functions for interrupt control.
14825 * config/nds32/nds32-intrinsic.md: Likewise.
14826 * config/nds32/nds32_intrinsic.h: Likewise.
14827 * config/nds32/nds32.h (nds32_builtins): Likewise.
14829 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
14831 * config/nds32/nds32.c (nds32_init_machine_status,
14832 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
14833 strict_aligned_p field.
14834 (nds32_expand_to_rtl_hook): New function.
14835 (TARGET_EXPAND_TO_RTL_HOOK): Define.
14836 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
14838 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14839 Chung-Ju Wu <jasonwucj@gmail.com>
14841 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
14842 * config/nds32/nds32-n7.md: New file.
14843 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
14844 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
14846 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
14847 * config/nds32/nds32.md (pipeline_model): Add n7.
14848 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
14849 * config/nds32/pipelines.md: Include n7 settings.
14851 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14852 Chung-Ju Wu <jasonwucj@gmail.com>
14854 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
14855 * config/nds32/nds32-e8.md: New file.
14856 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
14857 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
14859 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
14860 * config/nds32/nds32.md (pipeline_model): Add e8.
14861 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
14862 * config/nds32/pipelines.md: Include e8 settings.
14864 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14865 Chung-Ju Wu <jasonwucj@gmail.com>
14867 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
14868 * config/nds32/nds32-n8.md: New file.
14869 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
14870 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
14872 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
14873 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
14874 * config/nds32/nds32.md (pipeline_model): Add n8.
14875 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
14876 * config/nds32/pipelines.md: Include n8 settings.
14878 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14879 Chung-Ju Wu <jasonwucj@gmail.com>
14881 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
14882 * config/nds32/nds32-n9-2r1w.md: New file.
14883 * config/nds32/nds32-n9-3r2w.md: New file.
14884 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
14885 nds32_register_ports): New or modify for cpu n9.
14886 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
14888 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
14889 * config/nds32/nds32-utils.c: New file.
14890 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
14891 TARGET_MUL_SLOW): Define.
14892 * config/nds32/nds32.md (pipeline_model): New attribute.
14893 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
14894 New options that support cpu n9.
14895 * config/nds32/pipelines.md: Include n9 settings.
14896 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
14898 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
14900 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
14901 information if necessary.
14902 (output_cond_branch_compare_zero): Likewise.
14903 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
14904 (nds32_target_alignment): Refine for alignment.
14905 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
14906 (FUNCTION_BOUNDARY): Modify.
14907 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
14909 * config/nds32/nds32.opt (malways-align, malign-functions): New.
14911 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
14913 * config/nds32/constants.md (unspec_volatile_element): Add values for
14914 TLB operation and data prefetch.
14915 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
14916 functions for TLB operation and data prefetch.
14917 * config/nds32/nds32-intrinsic.md: Likewise.
14918 * config/nds32/nds32_intrinsic.h: Likewise.
14919 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
14920 (nds32_print_operand): Likewise.
14921 * config/nds32/nds32.h (nds32_builtins): Likewise.
14923 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
14924 Andrew Pinski <pinsika@gcc.gnu.org>
14926 PR middle-end/82976
14927 * match.pd: Use constant_boolean_node of correct type instead of
14928 boolean_true_node or boolean_false_node for simplifying
14929 pointer comparisons to zero.
14931 2018-04-07 Jakub Jelinek <jakub@redhat.com>
14933 PR tree-optimization/80021
14934 * tree.c (verify_type_variant): Make error call in verify_variant_match
14935 translatable and remove final full stop.
14937 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14939 * config/nds32/constants.md (unspec_volatile_element): Add
14940 UNSPEC_VOLATILE_EH_RETURN.
14941 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
14942 nds32_output_stack_pop): Support dwarf exception handling process.
14943 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
14944 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
14945 exception handling process.
14946 (nds32_compute_stack_frame): Likewise.
14947 (nds32_return_addr_rtx): Likewise.
14948 (nds32_initial_elimination_offset): Likewise.
14949 (nds32_expand_prologue): Likewise.
14950 (nds32_expand_epilogue): Likewise.
14951 (nds32_dynamic_chain_address): New function.
14952 * config/nds32/nds32.h (machine_function): Add fields for dwarf
14953 exception handling.
14954 (DYNAMIC_CHAIN_ADDRESS): Define.
14955 (EH_RETURN_DATA_REGNO): Define.
14956 (EH_RETURN_STACKADJ_RTX): Define.
14957 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
14958 patterns for dwarf exception handling.
14960 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14962 * config/nds32/nds32.h: Clean up obsolete macros.
14964 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14966 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
14967 Add enum values for particular instructions.
14968 * config/nds32/nds32-intrinsic.c: Implementation of expanding
14969 particular intrinsic functions.
14970 * config/nds32/nds32-intrinsic.md: Likewise.
14971 * config/nds32/nds32_intrinsic.h: Likewise.
14972 * config/nds32/nds32.h (nds32_builtins): Likewise.
14973 * config/nds32/nds32.md (type): Add pbsad and pbsada.
14974 (btst, ave): New patterns for particular instructions.
14976 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14978 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
14979 Add enum values for atomic load/store and memory sync.
14980 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
14982 * config/nds32/nds32-intrinsic.md: Likewise.
14983 * config/nds32/nds32_intrinsic.h: Likewise.
14984 * config/nds32/nds32.h (nds32_builtins): Likewise.
14986 2018-04-07 Jakub Jelinek <jakub@redhat.com>
14988 PR tree-optimization/85257
14989 * fold-const.c (native_encode_vector): If not all elts could fit
14990 and off is -1, return 0 rather than offset.
14991 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
14992 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
14993 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
14994 adjust buffer in native_interpret_expr call.
14996 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14998 * config/nds32/constants.md (unspec_volatile_element): Add cache
14999 control enum values.
15000 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
15001 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
15002 * config/nds32/nds32.c (nds32_cctl_names): New.
15003 (nds32_print_operand): Handle cache control register names.
15004 * config/nds32/nds32.h (nds32_builtins): New enum values.
15005 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
15007 * config/nds32/nds32.md (type): Add mmu.
15008 * config/nds32/pipelines.md (simple_insn): Add mmu.
15010 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15012 * config/nds32/nds32.md (type): Remove call.
15013 * config/nds32/pipelines.md (simple_insn): Likewise.
15015 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15017 * config/nds32/constants.md (unspec_volatile_element): Add
15018 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
15019 UNSPEC_VOLATILE_FMFCFG.
15020 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
15021 description for fmfcfg and fmfcsr.
15022 (bdesc_1arg): Add fmtcsr.
15023 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
15024 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
15025 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
15026 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
15027 unspec_fmfcfg): New patterns.
15028 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
15029 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
15030 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
15031 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
15032 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
15033 __nds32__fmfcfg): Define.
15035 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15037 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
15038 intrinsic register names.
15039 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
15040 intrinsic register enum values and macros.
15042 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15044 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
15045 for load/store addressing form.
15046 (nds32_print_operand_address): Likewise.
15048 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
15051 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
15052 based on LABEL_REF. Remove useless assertion.
15053 (pic_address_needs_scratch): Fix formatting.
15054 (sparc_legitimize_pic_address): Minor tweaks.
15055 (sparc_delegitimize_address): Adjust assertion accordingly.
15056 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
15057 into symbolic_operand.
15058 (movsi_high_pic_label_ref): Likewise.
15059 (movsi_lo_sum_pic_label_ref): Likewise.
15060 (movdi_pic_label_ref): Likewise.
15061 (movdi_high_pic_label_ref): Likewise.
15062 (movdi_lo_sum_pic_label_ref): Likewise.
15064 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
15066 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
15067 custom LIB_SPEC setup.
15069 2018-04-06 Ruslan Bukin <br@bsdpad.com>
15070 Kito Cheng <kito.cheng@gmail.com>
15072 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
15073 * config/riscv/freebsd.h: New.
15075 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
15077 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
15078 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
15081 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
15082 Kito Cheng <kito.cheng@gmail.com>
15084 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
15085 nds32_output_call, nds32_symbol_binds_local_p): New functions.
15086 * config/nds32/nds32-protos.h (nds32_output_call,
15087 nds32_output_return): Declare.
15088 * config/nds32/nds32.md: Refine all the call and return patterns.
15090 2018-04-06 Jakub Jelinek <jakub@redhat.com>
15093 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
15094 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
15096 PR rtl-optimization/84872
15097 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
15098 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
15099 EDGE_CROSSING edge.
15101 2018-04-06 Tamar Christina <tamar.christina@arm.com>
15103 * expr.c (copy_blkmode_to_reg): Revert 254862.
15104 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
15106 2018-04-06 Richard Biener <rguenther@suse.de>
15108 PR middle-end/85244
15109 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
15110 after seeing a component reference with an adjacent field. Treat
15111 refs to arrays at struct end of external decls similar to
15112 refs to unconstrained commons.
15114 2018-04-06 Jakub Jelinek <jakub@redhat.com>
15117 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
15118 look through SAVE_EXPRs with non-side-effects argument. Adjust
15120 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
15123 2018-04-06 Richard Biener <rguenther@suse.de>
15125 PR middle-end/85180
15126 * alias.c (find_base_term): New wrapper around find_base_term
15127 unwinding CSELIB_VAL_PTR changes.
15128 (find_base_term): Do not restore CSELIB_VAL_PTR during the
15131 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15133 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
15135 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
15136 constant definitions.
15137 ("nop"): lr 0,0 -> nopr r0
15138 ("nop_lr0", "nop_lr1"): New insn definitions.
15140 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
15142 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
15143 NDS32_V3PUSH_AVAILABLE_P macro.
15145 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
15146 Chung-Ju Wu <jasonwucj@gmail.com>
15148 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
15149 (nds32*-*-*): Add float and fpu_config into supported_defaults.
15150 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
15151 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
15152 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
15153 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
15154 * config/nds32/constraints.md: New constraints and checking for hard
15155 float configuration.
15156 * config/nds32/iterators.md: New mode iterator and attribute for hard
15157 float configuration.
15158 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
15160 * config/nds32/nds32-fpu.md: New file.
15161 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
15162 deal with hard float code generation.
15163 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
15165 (abi_type, float_reg_number): New enum type.
15166 * config/nds32/nds32-predicates.c: New predicates for hard float.
15167 * config/nds32/nds32-protos.h: Declare functions for hard float.
15168 * config/nds32/nds32.c: Implementation for hard float configuration.
15169 * config/nds32/nds32.h: Definitions for hard float configuration.
15170 * config/nds32/nds32.md: Include hard float machine description and
15171 modify patterns for hard float configuration.
15172 * config/nds32/nds32.opt: New options for hard float configuration.
15173 * config/nds32/predicates.md: New predicates for hard float
15176 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
15178 * common/config/nds32/nds32-common.c
15179 (nds32_option_optimization_table): Enable -mreleax-hint by default.
15181 2018-04-05 Jakub Jelinek <jakub@redhat.com>
15183 PR middle-end/85195
15184 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
15185 CONSTRUCTOR_ELT (ctor, ...)->value.
15187 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
15190 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
15192 2018-04-05 Tom de Vries <tom@codesourcery.com>
15195 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
15198 2018-04-05 Shiva Chen <shiva0217@gmail.com>
15199 Kito Cheng <kito.cheng@gmail.com>
15201 * config/nds32/constraints.md (U33): Fine-tune checking condition.
15202 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
15203 * config/nds32/nds32.h (nds32_16bit_address_type): Add
15204 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
15206 2018-04-05 Shiva Chen <shiva0217@gmail.com>
15207 Kito Cheng <kito.cheng@gmail.com>
15209 * config/nds32/constraints.md (Ufe): New memory constraint.
15210 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
15211 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
15212 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
15214 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
15215 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
15217 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15219 * config/nds32/nds32.md: Use optimize_size in the condition for
15220 alu-shift instructions.
15222 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15224 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
15226 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15228 * config/nds32/nds32.md (negsi2): Refine pattern.
15230 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
15231 Chung-Ju Wu <jasonwucj@gmail.com>
15233 * config/nds32/iterators.md (shift_rotate): New code iterator.
15234 (shift): New code attribute.
15235 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
15236 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
15237 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
15238 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
15239 bit-wise operations.
15240 (andsi3, *andsi3): Ditto.
15241 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
15242 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
15243 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
15244 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
15245 nds32_ior_operand, nds32_xor_operand): New predicates.
15247 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15249 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
15250 (addsi3, subsi3): ... this.
15252 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15254 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
15256 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15258 * config/nds32/nds32.md: Adjust indention.
15260 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
15262 * config/nds32/nds32.md (feature): New attribute.
15264 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15266 * config/nds32/nds32.md (subtype): New attribute.
15268 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
15271 * config/arm/arm-builtins.c (arm_expand_builtin): Change
15272 expansion to perform a bitwise AND of the argument followed by a
15273 boolean negation of the result.
15275 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
15277 PR rtl-optimization/84878
15278 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
15279 the basic block. Assert the use reference is not artificial and that
15280 it has an associated insn.
15282 2018-04-04 Michael Matz <matz@suse.de>
15284 * builtins.c (compute_objsize): Pass correct operand
15285 to array_at_struct_end_p.
15287 2018-04-04 Richard Biener <rguenther@suse.de>
15290 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
15291 from contexts for DINFO_LEVEL_TERSE and below.
15293 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15295 * config/nds32/nds32-doubleword.md (move_<mode>): Require
15296 resiter_operand condition.
15297 * config/nds32/nds32.md (*move<mode>): Ditto.
15299 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15300 Monk Chiang <sh.chiang04@gmail.com>
15302 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
15304 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15306 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
15308 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15309 Kito Cheng <kito.cheng@gmail.com>
15311 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
15312 nds32_cond_code_str, output_cond_branch,
15313 output_cond_branch_compare_zero, nds32_expand_cbranch,
15314 nds32_expand_cstore, nds32_expand_movcc,
15315 nds32_output_cbranchsi4_equality_zero,
15316 nds32_output_cbranchsi4_equality_reg,
15317 nds32_output_cbranchsi4_equality_reg_or_const_int,
15318 nds32_output_cbranchsi4_greater_less_zero: New functions.
15319 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
15320 nds32_expand_cstore, nds32_expand_movcc,
15321 nds32_output_cbranchsi4_equality_zero,
15322 nds32_output_cbranchsi4_equality_reg,
15323 nds32_output_cbranchsi4_equality_reg_or_const_int,
15324 nds32_output_cbranchsi4_greater_less_zero): Declare.
15325 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
15326 nds32_rimm11s_operand): New predicates.
15327 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
15328 * config/nds32/nds32.md: Rewrite all the branch and conditional move
15331 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15333 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
15334 * config/nds32/nds32.md: Ditto.
15335 * config/nds32/pipelines.md: Ditto.
15337 2018-04-04 Richard Biener <rguenther@suse.de>
15339 PR tree-optimization/85168
15340 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
15341 propagating abnormals.
15343 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15345 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
15347 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15348 Kito Cheng <kito.cheng@gmail.com>
15350 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
15351 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
15352 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
15353 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
15354 * config/nds32/nds32.md (sibcall_internal): New.
15355 (sibcall_register): Remove.
15356 (sibcall_immediate): Remove.
15357 (sibcall_value_internal): New.
15358 (sibcall_value_register): Remove.
15359 (sibcall_value_immediate): Remove.
15360 * config/nds32/predicates.md (nds32_general_register_operand): New.
15361 (nds32_call_address_operand): New.
15363 2018-04-03 Jakub Jelinek <jakub@redhat.com>
15365 PR rtl-optimization/85167
15366 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
15367 bb_defs if *split_p, instead preinitialize it to NULL.
15369 PR tree-optimization/85156
15370 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
15371 evaluating the argument multiple times.
15373 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
15375 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
15377 (_mm_cvtpd_ps): Likewise.
15378 (_mm_cvttpd_epi32): Likewise.
15379 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
15380 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
15381 vector, pixel, and bool following altivec.h include.
15383 2018-04-03 Martin Sebor <msebor@redhat.com>
15385 * doc/extend.texi (Common Function Attributes): Clarify.
15386 (const attribute): Likewise.
15387 (pure attribute): Likewise.
15389 2018-04-03 Jakub Jelinek <jakub@redhat.com>
15392 * config/i386/i386.c (ix86_expand_vector_set): Use
15393 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
15395 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
15397 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
15398 instructions when changing rounding bits to preserve precision bits
15399 in the x87 control word.
15401 2018-04-03 Martin Liska <mliska@suse.cz>
15403 PR tree-optimization/82491
15404 * rtl.h (strip_offset_and_add): Replace += suboffset with
15405 poly_uint64 () + suboffset.
15407 2018-03-29 Martin Liska <mliska@suse.cz>
15408 Martin Jambor <mjambor@suse.cz>
15411 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
15412 param_type is not an integral or pointer type.
15414 2018-04-03 Richard Biener <rguenther@suse.de>
15416 * sese.h (recompute_all_dominators): Remove.
15418 2018-04-02 Martin Sebor <msebor@redhat.com>
15420 * doc/invoke.texi (-Wrestrict): Fix typos.
15422 2018-04-02 Jim Wilson <jimw@sifive.com>
15424 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
15425 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
15426 (<optab>di3, <optab>si3_extend): Likewise.
15427 (<optab>si3_mask, <optab>si3_mask_1): New.
15428 (<optab>di3_mask, <optab>di3_mask_1): New.
15429 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
15430 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
15431 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
15433 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
15435 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
15438 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
15440 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
15441 (nds32_canonicalize_comparison): New function.
15443 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
15444 Kito Cheng <kito.cheng@gmail.com>
15445 Kuan-Lin Chen <kuanlinchentw@gmail.com>
15447 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
15448 * config/nds32/constants.md (unspec_volatile_element): Add
15449 UNSPEC_VOLATILE_RELAX_GROUP.
15450 * config/nds32/nds32-relax-opt.c: New file.
15451 * config/nds32/nds32-predicates.c
15452 (nds32_symbol_load_store_p): New function.
15453 * config/nds32/nds32-protos.h
15454 (nds32_symbol_load_store_p): Declare function.
15455 (make_pass_nds32_relax_opt): Declare new rtl pass function.
15456 * config/nds32/nds32.c
15457 (nds32_register_pass): New function to register pass.
15458 (nds32_register_passes): New function to register passes.
15459 * config/nds32/nds32.md (relax_group): New pattern.
15460 * config/nds32/nds32.opt (mrelax-hint): New option.
15461 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
15463 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
15465 * config/nds32/t-nds32: Modify files dependency.
15467 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
15469 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
15470 (PROFILE_HOOK): Define its implementation.
15472 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
15474 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
15475 type and 32-bit size.
15477 2018-04-01 Jakub Jelinek <jakub@redhat.com>
15479 PR middle-end/85090
15480 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
15481 (V_128_256): New mode iterator.
15482 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
15483 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
15484 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
15486 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
15487 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
15489 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
15492 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
15493 NaN inputs correctly.
15495 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
15498 * config/rs6000/vsx.md (??r): New mode attribute.
15499 (*vsx_mov<mode>_64bit): Use it.
15500 (*vsx_mov<mode>_32bit): Likewise.
15502 2018-03-30 Martin Sebor <msebor@redhat.com>
15504 PR tree-optimization/84818
15505 * builtins.c (check_access): Use warning_n.
15507 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15510 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
15512 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
15515 2018-03-30 Julia Koval <julia.koval@intel.com>
15518 * x86-tune.def (movx, partial_reg_dependency): Enable for
15521 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
15523 PR inline-asm/84985
15524 * lra-constraints.c (process_alt_operands): Move setting
15525 this_alternative_matches below.
15527 2018-03-29 Martin Liska <mliska@suse.cz>
15530 * doc/invoke.texi: Document how LTO works with debug info.
15531 Describe auto-load support of binutils. Mention 'x86-64'
15532 as valid option value of -march option.
15534 2018-03-29 Jakub Jelinek <jakub@redhat.com>
15536 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
15539 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
15540 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
15541 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
15544 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
15547 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
15548 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
15549 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
15550 for RS6000_BTM_POWERPC64.
15551 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
15552 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
15553 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
15556 (DIVDEU): Likewise.
15558 2018-03-28 Carl Love <cel@us.ibm.com>
15561 2017-09-27 Carl Love <cel@us.ibm.com>
15563 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
15564 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
15565 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
15568 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15570 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
15571 instead of __vector bool.
15572 (_mm_max_pu8): Likewise.
15573 (_mm_min_pi16): Likewise.
15575 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
15578 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
15579 (DIVWEUO): Likewise.
15580 (DIVDEO): Likewise.
15581 (DIVDEUO): Likewise.
15582 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
15583 DIVWEUO and DIVDEUO.
15584 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
15585 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
15586 (div_extend): Likewise.
15587 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
15589 (__builtin_divweuo): Likewise.
15590 (__builtin_divdeo): Likewise.
15591 (__builtin_divdeuo): Likewise.
15593 2018-03-28 Jakub Jelinek <jakub@redhat.com>
15596 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
15597 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
15599 PR tree-optimization/82004
15600 * gimple-match-head.c (optimize_pow_to_exp): New function.
15601 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
15602 Don't fold to exp if optimize_pow_to_exp is false.
15604 2018-03-28 Martin Liska <mliska@suse.cz>
15607 * calls.c (initialize_argument_information): Fix trailing space.
15608 * common.opt: Fix typo and provide better explanation for
15609 -fsanitize-coverage option.
15610 * config/i386/i386.opt: Fix typo.
15612 2018-03-28 Jakub Jelinek <jakub@redhat.com>
15613 Martin Liska <mliska@suse.cz>
15616 * gimplify.c (asan_poison_variable): Don't do the check for
15617 gimplify_omp_ctxp here.
15618 (gimplify_decl_expr): Do it here.
15619 (gimplify_target_expr): Likewise.
15621 2018-03-28 Martin Liska <mliska@suse.cz>
15624 * config/i386/i386.c (ix86_function_arg_advance): Do not call
15625 chkp_type_bounds_count if MPX is not enabled.
15627 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
15629 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
15631 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
15634 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
15635 function to create the function decl for complex long double
15636 multiply and divide for -mabi=ieeelongdouble.
15637 (init_float128_ieee): Call it.
15639 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
15642 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
15643 -fcf-protection=branch -mibt.
15644 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
15646 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
15649 * config/arm/arm.c (arm_valid_symbolic_address): Handle
15650 arm_word_relocations.
15652 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
15655 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
15656 extern array declarations.
15658 2018-03-27 Richard Biener <rguenther@suse.de>
15660 PR middle-end/84067
15661 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
15662 explicit single_use checks.
15664 2018-03-27 Richard Biener <rguenther@suse.de>
15666 PR tree-optimization/85082
15667 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
15670 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
15672 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
15673 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
15674 Turn on fasynchronous-unwind-tables and funwind-tables.
15676 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
15679 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
15680 (*bmi_blsr_<mode>_ccz): Ditto.
15682 2018-03-26 Tom de Vries <tom@codesourcery.com>
15684 PR tree-optimization/85063
15685 * omp-general.c (offloading_function_p): New function. Factor out
15687 * omp-offload.c (pass_omp_target_link::gate): ... here.
15688 * omp-general.h (offloading_function_p): Declare.
15689 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
15690 with attribute omp declare target for offloading functions.
15692 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
15694 PR tree-optimization/84005
15695 * tree-data-ref.h (get_base_for_alignment): Declare.
15696 * tree-data-ref.c (get_base_for_alignment_1): New function.
15697 (get_base_for_alignment): Likewise.
15698 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
15699 get_base_for_alignment to find a suitable base object, instead
15700 of always using drb->base_address.
15702 2018-03-23 Jakub Jelinek <jakub@redhat.com>
15704 PR inline-asm/85022
15705 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
15706 known size by default.
15708 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
15710 PR inline-asm/85030
15711 * lra-constraints.c (process_alt_operands): Don't match BLKmode
15712 and non BLKmode operands.
15714 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15717 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
15718 Clean up attributes.
15720 2018-03-23 Richard Biener <rguenther@suse.de>
15723 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
15724 we are going to emit early debug for LTO.
15726 2018-03-23 Jakub Jelinek <jakub@redhat.com>
15728 PR inline-asm/85034
15729 * function.c (match_asm_constraints_1): Don't optimize if input
15730 doesn't satisfy general_operand predicate for output's mode.
15732 PR inline-asm/85022
15733 * alias.c (write_dependence_p): Don't require for x_canonicalized
15734 non-VOIDmode if x has VOIDmode.
15737 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
15738 just don't try to optimize it rather than assert it never happens.
15740 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
15742 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
15743 macro expansions for definition of ST_INTERNAL_<mode> and
15744 LD_INTERNAL_<mode> builtins.
15745 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
15747 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
15749 (altivec_expand_st_builtin): Likewise.
15750 (altivec_expand_builtin): Remove calls to deleted functions.
15751 (rs6000_address_for_altivec): Delete this function.
15752 * config/rs6000/vector.md: Remove expands for
15753 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
15755 2018-03-22 Sudakshina Das <sudi.das@arm.com>
15758 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
15759 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
15760 re-computing once computed.
15761 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
15762 (arm_init_machine_status): Initialize
15763 machine->static_chain_stack_bytes.
15765 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
15768 * doc/extend.texi: Add four new prototypes for vec_ld.
15769 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
15770 definitions for more logical presentation.
15771 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
15772 entries for V1TI variants of __builtin_altivec_ld builtin.
15773 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
15774 handling of V1TI variant of LVX icode pattern.
15775 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
15776 (rs6000_gimple_fold_builtin): Likewise.
15777 (altivec_init_builtins): Add code to define
15778 __builtin_altivec_lvx_v1ti function.
15780 2018-03-22 Jakub Jelinek <jakub@redhat.com>
15782 PR inline-asm/84941
15783 * function.c (match_asm_constraints_1): Don't do the optimization
15784 if input isn't a REG, SUBREG, MEM or constant.
15786 2018-03-22 Tom de Vries <tom@codesourcery.com>
15788 PR tree-optimization/84956
15789 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
15790 bb_has_abnormal_pred.
15792 2018-03-22 Jakub Jelinek <jakub@redhat.com>
15795 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
15796 DECL_INITIAL (decl) to decl at the end.
15797 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
15798 adjust the comment.
15800 2018-03-21 Joseph Myers <joseph@codesourcery.com>
15802 * doc/extend.texi (__builtin_tgmath): Document when complex
15803 integer types are treated as _Complex _Float64.
15805 2018-03-21 Tom de Vries <tom@codesourcery.com>
15807 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
15809 2018-03-21 Jakub Jelinek <jakub@redhat.com>
15811 PR tree-optimization/84960
15812 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
15813 if it is ENTRY block, move them into single succ of ENTRY in that case.
15815 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
15817 PR tree-optimization/84811
15818 * poly-int.h (poly_span_traits): Remove the T3 parameter and
15819 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
15820 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
15821 (known_subrange_p): Update accordingly. Cast each value involved
15822 in the size comparison, rather than casting the result of the
15825 2018-03-21 Jakub Jelinek <jakub@redhat.com>
15827 PR tree-optimization/84982
15828 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
15829 by flipping the least significant bit rather than all bits from
15830 bitpos to bitpos + bitsize - 1.
15832 2018-03-21 Nathan Sidwell <nathan@acm.org>
15834 * doc/extend.texi (Deprecated Features): Remove mention of
15835 long-deleted deprecations.
15837 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15840 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
15841 * configure: Regenerate.
15843 2018-03-21 Tom de Vries <tom@codesourcery.com>
15845 PR tree-optimization/83126
15846 * tree-parloops.c (num_phis): New function.
15847 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
15849 2018-03-21 Nathan Sidwell <nathan@acm.org>
15851 * doc/extend.texi (Deprecated Features): Update deprecated flags,
15852 mention anon-struct/union members and trailing attributes.
15854 2018-03-21 Bin Cheng <bin.cheng@arm.com>
15856 PR tree-optimization/84969
15857 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
15858 builtin memset partitions if they set different rhs values.
15860 2018-03-21 Jakub Jelinek <jakub@redhat.com>
15862 PR rtl-optimization/84989
15863 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
15864 VEC_DUPLICATE with scalar result mode.
15866 2018-03-21 Martin Liska <mliska@suse.cz>
15869 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
15870 not intended return statement.
15872 2018-03-21 Martin Liska <mliska@suse.cz>
15875 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
15876 (chkp_find_bound_slots_1): Limit number of iterations.
15878 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
15881 * Minor grammar fixes for x86 options.
15883 2018-03-20 Jakub Jelinek <jakub@redhat.com>
15886 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
15887 holding REG_CFA_RESTORE notes, instead turn them into a USE.
15889 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
15892 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
15893 (altivec_lvx_<mode>_1op): Likewise.
15894 (altivec_stvx_<mode>_2op): Likewise.
15895 (altivec_stvx_<mode>_1op): Likewise.
15896 (altivec_lvx_<VM2:mode>): New define_expand.
15897 (altivec_stvx_<VM2:mode>): Likewise.
15898 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
15899 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
15900 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
15901 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
15902 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
15903 (rs6000_gen_lvx): Likewise.
15904 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
15905 (altivec_expand_stv_builtin): Likewise.
15906 (altivec_expand_builtin): Likewise.
15907 * config/rs6000/vector.md: Likewise.
15909 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15912 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
15915 2018-03-20 Richard Biener <rguenther@suse.de>
15918 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
15919 sign-conversions as zero, fall back to standard scalar_stmt
15922 2018-03-20 Martin Liska <mliska@suse.cz>
15925 * predict.c (rebuild_frequencies): Handle case when we have
15926 PROFILE_ABSENT, but flag_guess_branch_prob is false.
15928 2018-03-20 Jakub Jelinek <jakub@redhat.com>
15931 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
15932 flag_section_anchors.
15933 * varasm.c (use_blocks_for_decl_p): Remove hack for
15934 dw2_force_const_mem.
15937 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
15939 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
15940 be created, use lowpart_subreg of operands[0] rather than operands[0]
15942 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
15943 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
15944 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
15945 and n constraint instead of aarch64_shift_imm_di and Usd.
15946 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
15947 (*aarch64_<optab>_reg_minus<mode>3): ... this.
15949 2018-03-20 Sudakshina Das <sudi.das@arm.com>
15952 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
15953 to favor GPR over NEON registers.
15954 (<shift>di3_neon): Likewise.
15956 2018-03-20 Tom de Vries <tom@codesourcery.com>
15959 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
15960 (nvptx_process_pars): Emit bar.sync asap and alap.
15962 2018-03-20 Tom de Vries <tom@codesourcery.com>
15965 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
15966 seen_label if seen_label is already set.
15968 2018-03-20 Jakub Jelinek <jakub@redhat.com>
15971 * config/i386/i386.c (fold_builtin_cpu): For features above 31
15972 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
15973 Use 1U instead of 1. Formatting fixes.
15976 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
15977 instead of TREE_TYPE (s1) for the return value.
15979 2018-03-19 Jakub Jelinek <jakub@redhat.com>
15981 PR tree-optimization/84946
15982 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
15983 bitsize + bitsize in poly_uint64 rather than poly_int64.
15986 * dwarf2asm.c: Include fold-const.c.
15987 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
15988 of decl rather than decl itself.
15990 PR rtl-optimization/84643
15991 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
15993 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
15996 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
15997 calling assemble_variable.
15999 2018-03-19 Sudakshina Das <sudi.das@arm.com>
16002 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
16003 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
16005 2018-03-19 Jim Wilson <jimw@sifive.com>
16008 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
16009 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
16010 (riscv_first_stack_step): Likewise.
16011 (riscv_option_override): Use STACK_BOUNDARY instead of
16012 MIN_STACK_BOUNDARY.
16013 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
16014 MIN_STACK_BOUNDARY.
16015 (BIGGEST_ALIGNMENT): Set to 128.
16016 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
16017 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
16020 2018-03-19 Richard Biener <rguenther@suse.de>
16022 PR tree-optimization/84933
16023 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
16024 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
16026 2018-03-19 Richard Biener <rguenther@suse.de>
16028 PR tree-optimization/84859
16029 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
16030 (cond_if_else_store_replacement): Perform sinking operation on
16031 single-store BBs regardless of MAX_STORES_TO_SINK setting.
16032 Generalize what a BB with a single eligible store is.
16034 2018-03-19 Richard Biener <rguenther@suse.de>
16036 PR tree-optimization/84929
16037 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
16038 chrec_is_positive against non-chrec arg.
16040 2018-03-19 Tamar Christina <tamar.christina@arm.com>
16043 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
16045 2018-03-18 Martin Liska <mliska@suse.cz>
16047 PR rtl-optimization/84635
16048 * regrename.c (build_def_use): Use matches_mode only when
16051 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
16053 PR tree-optimization/84913
16054 * tree-vect-loop.c (vectorizable_reduction): Don't try to
16055 vectorize chains of COND_EXPRs.
16057 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
16059 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
16061 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
16063 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
16065 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
16067 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
16069 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
16070 Kito Cheng <kito.cheng@gmail.com>
16072 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
16073 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
16074 (nds32_adjust_reg_alloc_order): New function.
16075 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
16077 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
16079 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
16080 nds32_print_operand, nds32_print_operand_address): Use
16081 HOST_WIDE_INT_PRINT_DEC instead.
16083 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
16085 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
16087 2018-03-17 Jakub Jelinek <jakub@redhat.com>
16090 * config/i386/i386.c (initial_ix86_tune_features,
16091 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
16092 unsigned long long.
16093 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
16094 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
16095 rather than 1u << ix86_tune. Formatting fix.
16096 (ix86_option_override_internal): Change ix86_arch_mask from
16097 unsigned int to unsigned HOST_WIDE_INT, initialize to
16098 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
16099 (ix86_function_specific_restore): Likewise.
16101 2018-03-16 Jakub Jelinek <jakub@redhat.com>
16104 * postreload.c (reload_combine_recognize_pattern): Perform
16105 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
16106 truncate_int_for_mode the result for the destination's mode.
16109 * hsa-gen.c (mem_type_for_type): Fix comment typo.
16110 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
16112 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
16115 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
16118 * lra-assigns.c (lra_split_hard_reg_for): Don't use
16119 regno_allocno_class_array and sorted_pseudos.
16120 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
16121 insns where regno is used.
16123 2018-03-16 Martin Liska <mliska@suse.cz>
16126 * multiple_target.c (create_dispatcher_calls): Redirect
16127 reference in the symbol table.
16129 2018-03-16 Martin Liska <mliska@suse.cz>
16132 * multiple_target.c (create_dispatcher_calls): Redirect also
16135 2018-03-16 Jakub Jelinek <jakub@redhat.com>
16139 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
16140 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
16141 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
16143 2018-03-16 Julia Koval <julia.koval@intel.com>
16145 * doc/invoke.texi (Skylake Server): Add CLWB.
16146 Cannonlake): Remove CLWB.
16148 2018-03-16 Jakub Jelinek <jakub@redhat.com>
16150 PR tree-optimization/84841
16151 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
16153 (FLOAT_ONE_CONST_TYPE): Define.
16154 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
16155 (sort_by_operand_rank): Put entries with higher constant_type last
16156 rather than first to match comments.
16158 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
16160 * config/nios2/nios2.md (movsi_internal): Fix thinko in
16163 2018-03-15 Jakub Jelinek <jakub@redhat.com>
16166 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
16167 check and use address of target always.
16169 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
16172 * config/i386/i386.c (indirect_thunk_needed): Update comments.
16173 (indirect_thunk_bnd_needed): Likewise.
16174 (indirect_thunks_used): Likewise.
16175 (indirect_thunks_bnd_used): Likewise.
16176 (indirect_return_needed): New.
16177 (indirect_return_bnd_needed): Likewise.
16178 (output_indirect_thunk_function): Add a bool argument for
16180 (output_indirect_thunk_function): Don't generate alias for
16181 function return thunk.
16182 (ix86_code_end): Call output_indirect_thunk_function to generate
16183 function return thunks.
16184 (ix86_output_function_return): Set indirect_return_bnd_needed
16185 and indirect_return_needed instead of indirect_thunk_bnd_needed
16186 and indirect_thunk_needed.
16188 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
16190 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
16191 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
16192 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
16194 2018-03-15 David Malcolm <dmalcolm@redhat.com>
16195 Paul Hua <paul.hua.gm@gmail.com>
16198 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
16200 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
16202 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
16203 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
16204 resp. SFmode cases.
16206 2018-03-15 Tamar Christina <tamar.christina@arm.com>
16209 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
16210 instead of GET_MODE_SIZE when comparing Units.
16212 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
16215 * varasm.c (hash_section): Return an unchangeble hash value
16216 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
16217 Return !aarch64_can_use_per_function_literal_pools_p ().
16219 2018-03-15 Jakub Jelinek <jakub@redhat.com>
16222 * optabs.c (emit_conditional_move): Pass address of cmode's copy
16223 rather than address of cmode as last argument to prepare_cmp_insn.
16225 2018-03-15 Julia Koval <julia.koval@intel.com>
16227 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
16228 F_AVX512VNNI, F_AVX512BITALG): New.
16230 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
16233 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
16234 insn for floating-point loads and stores.
16236 2018-03-14 Carl Love <cel@us.ibm.com>
16238 * config/rs6000/rs6000-c.c: Add macro definitions for
16239 ALTIVEC_BUILTIN_VEC_PERMXOR.
16240 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
16241 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
16242 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
16243 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
16245 * config/doc/extend.texi: Add prototypes for vec_permxor.
16247 2018-03-14 David Malcolm <dmalcolm@redhat.com>
16250 * diagnostic-show-locus.c (class layout_point): Convert m_line
16251 from int to linenum_type.
16252 (line_span::comparator): Use linenum "compare" function when
16253 comparing line numbers.
16254 (test_line_span): New function.
16255 (layout_range::contains_point): Convert param "row" from int to
16257 (layout_range::intersects_line_p): Likewise.
16258 (layout::will_show_line_p): Likewise.
16259 (layout::print_source_line): Likewise.
16260 (layout::should_print_annotation_line_p): Likewise.
16261 (layout::print_annotation_line): Likewise.
16262 (layout::print_leading_fixits): Likewise.
16263 (layout::annotation_line_showed_range_p): Likewise.
16264 (struct line_corrections): Likewise for field m_row.
16265 (line_corrections::line_corrections): Likewise for param "row".
16266 (layout::print_trailing_fixits): Likewise.
16267 (layout::get_state_at_point): Likewise.
16268 (layout::get_x_bound_for_row): Likewise.
16269 (layout::print_line): Likewise.
16270 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
16271 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
16272 * input.c (selftest::test_linenum_comparisons): New function.
16273 (selftest::input_c_tests): Call it.
16274 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
16275 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
16276 * selftest.h (ASSERT_GT): New macro.
16277 (ASSERT_GT_AT): New macro.
16278 (ASSERT_LT): New macro.
16279 (ASSERT_LT_AT): New macro.
16281 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
16283 PR rtl-optimization/84780
16284 * combine.c (distribute_links): Don't make a link based on pc_rtx.
16286 2018-03-14 Martin Liska <mliska@suse.cz>
16288 * tree.c (record_node_allocation_statistics): Use
16289 get_stats_node_kind.
16290 (get_stats_node_kind): New function extracted from
16291 record_node_allocation_statistics.
16292 (free_node): Use get_stats_node_kind.
16294 2018-03-14 Richard Biener <rguenther@suse.de>
16296 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
16297 that the value-set of ANTIC_IN doesn't grow.
16300 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
16302 (BB_VISITED_WITH_VISITED_SUCCS): New define.
16303 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
16305 2018-03-14 Julia Koval <julia.koval@intel.com>
16307 * config.gcc (icelake-client, icelake-server): New.
16309 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
16310 (initial_ix86_arch_features): Ditto.
16311 (PTA_SKYLAKE): Add SGX.
16312 (PTA_ICELAKE): Remove.
16313 (PTA_ICELAKE_CLIENT): New.
16314 (PTA_ICELAKE_SERVER): New.
16315 (ix86_option_override_internal): Split up icelake on icelake client and
16317 (get_builtin_code_for_version): Ditto.
16318 (fold_builtin_cpu): Ditto.
16319 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
16320 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
16321 * config/i386/i386.h (processor_type): Ditto.
16322 * doc/invoke.texi: Ditto.
16324 2018-03-14 Jakub Jelinek <jakub@redhat.com>
16327 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
16328 INTEGER_CST offset, add it together with bitpos / 8 and
16329 sign extend based on POINTER_SIZE.
16333 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
16336 * config/i386/constraints.md (Yc): New register constraint.
16337 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
16338 Use Yc constraint for alternative 2 of operand 0. Remove
16339 preferred_for_speed attribute.
16341 2018-03-14 Richard Biener <rguenther@suse.de>
16343 PR tree-optimization/84830
16344 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
16345 with the old one to avoid oscillations.
16347 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
16350 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
16352 (assign_by_spills): Return a flag of reload assignment failure.
16353 Do not process the reload assignment failures. Do not spill other
16354 reload pseudos if they has the same reg class. Update n if
16356 (lra_assign): Add a return arg. Set up from the result of
16357 assign_by_spills call.
16358 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
16359 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
16360 usage_insns if it is not NULL.
16361 (spill_hard_reg_in_range): New function.
16362 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
16363 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
16364 function prototypes.
16365 (lra_assign): Change prototype.
16366 * lra.c (lra): Add code to deal with fails by splitting hard reg
16369 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
16371 * config/riscv/riscv.opt (mrelax): New option.
16372 * config/riscv/riscv.c (riscv_file_start): Emit ".option
16373 "norelax" when riscv_mrelax is disabled.
16374 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
16376 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16379 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
16380 reassociation for int modes.
16382 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
16384 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
16385 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
16387 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
16388 * config/aarch64/aarch64-sve.md
16389 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
16390 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
16391 (*extend<mode><Vwide>2): Rename to...
16392 (aarch64_sve_extend<mode><Vwide>2): ...this.
16393 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
16394 renaming the old pattern to...
16395 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
16397 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
16398 define_expand, renaming the old pattern to...
16399 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
16400 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
16401 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
16402 account when deciding which SVE instruction the optab should use.
16403 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
16405 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
16407 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
16408 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
16409 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
16410 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
16411 (tlsdesc_small_<mode>): Turn a define_expand and use
16412 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
16413 (tlsdesc_small_advsimd_<mode>): ...this.
16414 (tlsdesc_small_sve_<mode>): New pattern.
16416 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
16418 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
16419 (UNSPEC_UMUL_HIGHPART): New constants.
16420 (MUL_HIGHPART): New int iteraor.
16421 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
16422 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
16424 (*<su>mul<mode>3_highpart): New define_insn.
16426 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
16429 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
16432 2018-03-13 Martin Liska <mliska@suse.cz>
16435 * (sem_item_optimizer::sem_item_optimizer): Initialize new
16437 (sem_item_optimizer::~sem_item_optimizer): Release it.
16438 (sem_item_optimizer::merge_classes): Register variable aliases.
16439 (sem_item_optimizer::fixup_pt_set): New function.
16440 (sem_item_optimizer::fixup_points_to_sets): Likewise.
16441 * ipa-icf.h: Declare new variables and functions.
16443 2018-03-13 Jakub Jelinek <jakub@redhat.com>
16445 PR middle-end/84834
16446 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
16447 integer_pow2p@2 and test integer_pow2p in condition.
16448 (A < 0 ? C : 0): Similarly for @1.
16450 PR middle-end/84831
16451 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
16452 characters starting at p contain '\0' character, don't look beyond
16456 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
16457 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
16460 * reg-stack.c (change_stack): Change update_end var from int to
16461 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
16462 also call set_block_for_insn on the newly added insns and rescan.
16465 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
16466 on the last operand.
16469 * tree.c (stabilize_reference_1): Return save_expr (e) for
16470 STATEMENT_LIST even if it doesn't have side-effects.
16472 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
16474 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
16476 2018-03-12 Renlin Li <renlin.li@arm.com>
16478 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
16479 aarch64_output_scalar_simd_mov_immediate.
16481 2018-03-12 Martin Sebor <msebor@redhat.com>
16483 PR tree-optimization/83456
16484 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
16485 for perfectly overlapping calls to memcpy.
16486 (gimple_fold_builtin_memory_chk): Same.
16487 (gimple_fold_builtin_strcpy): Handle no-warning.
16488 (gimple_fold_builtin_stxcpy_chk): Same.
16489 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
16491 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
16493 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
16494 parameter. Use it for SFmode.
16495 (rs6000_function_arg_advance_1): Adjust.
16496 (rs6000_function_arg): Adjust.
16497 (rs6000_gimplify_va_arg): Pass false for that new parameter.
16499 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
16501 PR rtl-optimization/84169
16502 PR rtl-optimization/84780
16503 * combine.c (can_combine_p): Check for a 2-insn combination whether
16504 the destination register is used between the two insns, too.
16506 2018-03-12 Richard Biener <rguenther@suse.de>
16508 PR tree-optimization/84803
16509 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
16510 for refs DR analysis didn't process.
16512 2018-03-12 Richard Biener <rguenther@suse.de>
16514 PR tree-optimization/84777
16515 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
16516 force-vectorize loops ignore whether we are optimizing for size.
16518 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
16520 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
16521 (TARGET_MD_ASM_ADJUST): Define.
16523 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
16524 Kito Cheng <kito.cheng@gmail.com>
16525 Chung-Ju Wu <jasonwucj@gmail.com>
16527 * config/nds32/nds32.c (nds32_compute_stack_frame,
16528 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
16529 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
16530 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
16531 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
16532 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
16533 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
16534 * config/nds32/nds32.md (prologue, epilogue): Use macro
16535 NDS32_V3PUSH_AVAILABLE_P to do checking.
16537 2018-03-11 Jakub Jelinek <jakub@redhat.com>
16540 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
16541 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
16542 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
16543 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
16544 addition of most attributes on !orig_type_die or the attribute not
16545 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
16547 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
16548 Chung-Ju Wu <jasonwucj@gmail.com>
16550 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
16551 __NDS32_VH__ macro.
16552 * config/nds32/nds32.opt (mvh): New option.
16554 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
16555 Chung-Ju Wu <jasonwucj@gmail.com>
16557 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
16559 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
16560 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
16563 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
16564 Chung-Ju Wu <jasonwucj@gmail.com>
16566 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
16568 * config/nds32/nds32-multiple.md (strlensi): New pattern.
16569 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
16571 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
16572 Kito Cheng <kito.cheng@gmail.com>
16573 Chung-Ju Wu <jasonwucj@gmail.com>
16575 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
16576 UNSPEC_FFMISM and UNSPEC_FLMISM.
16577 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
16578 for ffb, ffmism and flmism.
16579 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
16580 (unspec_ffmism): Ditto.
16581 (unspec_flmism): Ditto.
16582 (nds32_expand_builtin_impl): Check if string extension is available.
16583 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
16584 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
16586 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
16589 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
16592 * lra-assigns.c (assign_by_spills): Return a flag of reload
16593 assignment failure. Do not process the reload assignment
16594 failures. Do not spill other reload pseudos if they has the same
16596 (lra_assign): Add a return arg. Set up from the result of
16597 assign_by_spills call.
16598 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
16599 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
16600 usage_insns if it is not NULL.
16601 (spill_hard_reg_in_range): New function.
16602 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
16603 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
16604 function prototypes.
16605 (lra_assign): Change prototype.
16606 * lra.c (lra): Add code to deal with fails by splitting hard reg
16609 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
16612 * config/i386/i386.opt: Replace Enforcment with Enforcement.
16614 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
16617 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
16618 (dw_val_node): Add val_symbolic_view.
16619 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
16620 (symview_upper_bound): New.
16621 (new_line_info_table): Initialize symviews_since_reset.
16622 (dwarf2out_source_line): Count symviews_since_reset and set
16623 symview_upper_bound.
16624 (dw_val_equal_p): Handle symview.
16625 (add_AT_symview): New.
16626 (print_dw_val): Handle symview.
16627 (attr_checksum, attr_checksum_ordered): Likewise.
16628 (same_dw_val_p, size_of_die): Likewise.
16629 (value_format, output_die): Likewise.
16630 (add_high_low_attributes): Use add_AT_symview for entry_view.
16631 (dwarf2out_finish): Reset symview_upper_bound, clear
16634 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
16637 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
16638 Add strict argument and use it.
16639 (rs6000_split_multireg_move): Update for new strict argument.
16640 (mem_operand_gpr): Disallow all non-offsettable addresses.
16641 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
16643 2018-03-09 Jakub Jelinek <jakub@redhat.com>
16646 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
16647 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
16648 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
16651 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
16652 decl, use remap_type if we want to use the type.
16654 2018-03-09 Martin Sebor <msebor@redhat.com>
16656 PR tree-optimization/84526
16657 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
16659 (builtin_access::generic_overlap): Be prepared to handle non-array
16662 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
16664 PR rtl-optimization/84682
16665 * lra-constraints.c (process_address_1): Check is_address flag
16666 for address constraints.
16667 (process_alt_operands): Likewise.
16668 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
16669 preprocess_constraints.
16670 * recog.h (preprocess_constraints): Add oploc parameter.
16672 * recog.c (preprocess_constraints): Test address_operand for
16673 CT_ADDRESS constraints.
16675 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
16678 * lra-assigns.c (assign_by_spills): Return a flag of reload
16679 assignment failure. Do not process the reload assignment
16680 failures. Do not spill other reload pseudos if they has the same
16682 (lra_assign): Add a return arg. Set up from the result of
16683 assign_by_spills call.
16684 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
16685 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
16686 usage_insns if it is not NULL.
16687 (spill_hard_reg_in_range): New function.
16688 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
16689 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
16690 function prototypes.
16691 (lra_assign): Change prototype.
16692 * lra.c (lra): Add code to deal with fails by splitting hard reg
16695 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16698 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
16699 Accept complain bool parameter. Only emit errors if it is true.
16700 (arm_parse_cpu_option_name): Likewise.
16701 (arm_target_thumb_only): Adjust callers of the above.
16702 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
16703 prototype to take a default true bool parameter.
16704 (arm_parse_arch_option_name): Likewise.
16706 2018-03-09 David Malcolm <dmalcolm@redhat.com>
16707 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
16711 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
16712 * configure: Regenerate.
16713 * configure.ac ("linker --version-script option"): New.
16714 ("linker soname option"): New.
16716 2018-03-09 Richard Biener <rguenther@suse.de>
16718 PR tree-optimization/84775
16719 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
16720 immediate uses of predicate stmts and mark them modified.
16723 PR tree-optimization/84178
16724 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
16726 (version_loop_for_if_conversion): Delay update_ssa call.
16727 (tree_if_conversion): Delay update_ssa until after predicate
16730 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
16733 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
16734 when the function accesses prior frames.
16736 2018-03-08 Jakub Jelinek <jakub@redhat.com>
16739 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
16740 gen_llsym, otherwise call maybe_gen_llsym.
16742 PR inline-asm/84742
16743 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
16744 has ',' character inside of it.
16746 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16749 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
16750 as clobbering CC_REGNUM.
16752 2018-03-08 Richard Biener <rguenther@suse.de>
16754 PR middle-end/84552
16755 * tree-scalar-evolution.c: Include tree-into-ssa.h.
16756 (follow_copies_to_constant): Do not follow SSA names registered
16759 2018-03-08 Richard Biener <rguenther@suse.de>
16761 PR tree-optimization/84178
16762 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
16764 (version_loop_for_if_conversion): Delay update_ssa call.
16765 (tree_if_conversion): Delay update_ssa until after predicate
16768 2018-03-08 David Malcolm <dmalcolm@redhat.com>
16770 PR tree-optimization/84178
16771 * tree-if-conv.c (release_bb_predicate): Remove the
16772 the assertion that the stmts have NULL use_ops.
16773 Discard the statements, asserting that they haven't
16774 yet been added to a BB.
16776 2018-03-08 Richard Biener <rguenther@suse.de>
16778 PR tree-optimization/84746
16779 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
16780 (phi_translate): Pass in destination ANTIC_OUT set.
16781 (phi_translate_1): Likewise. For a simplified result lookup
16782 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
16783 (phi_translate_set): Adjust.
16784 (do_pre_regular_insertion): Likewise.
16785 (do_pre_partial_partial_insertion): Likewise.
16787 2018-03-08 Martin Liska <mliska@suse.cz>
16789 PR gcov-profile/84735
16790 * doc/gcov.texi: Document usage of profile files.
16791 * gcov-io.h: Document changes in the format.
16793 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
16797 * dwarf2out.c (struct dw_line_info_table): Update comments for
16799 (FORCE_RESET_NEXT_VIEW): New.
16800 (FORCE_RESETTING_VIEW_P): New.
16801 (RESETTING_VIEW_P): Check for -1 too.
16802 (ZERO_VIEW_P): Likewise.
16803 (new_line_info_table): Force-reset next view.
16804 (dwarf2out_begin_function): Likewise.
16805 (dwarf2out_source_line): Simplify zero_view_p initialization.
16806 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
16807 view directly. Omit view when omitting .loc at line 0.
16809 2018-03-08 Jakub Jelinek <jakub@redhat.com>
16811 PR tree-optimization/84740
16812 * tree-switch-conversion.c (process_switch): Call build_constructors
16813 only if info.phi_count is non-zero.
16815 PR tree-optimization/84739
16816 * tree-tailcall.c (find_tail_calls): Check call arguments against
16817 DECL_ARGUMENTS (current_function_decl) rather than
16818 DECL_ARGUMENTS (func) when checking for tail recursion.
16820 2018-03-07 Jakub Jelinek <jakub@redhat.com>
16822 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
16823 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
16824 Volker Reichelt's entry and add entries for people that perform
16825 GCC fuzzy testing and report numerous bugs.
16827 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
16830 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
16831 readonly data in sdata, if that is disabled.
16832 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
16833 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
16834 -mreadonly-in-sdata option.
16836 2018-03-07 Martin Sebor <msebor@redhat.com>
16838 PR tree-optimization/84468
16839 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
16840 basic block when looking for nul assignment.
16842 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
16845 * except.h (output_function_exception_table): Adjust prototype.
16846 * except.c (output_function_exception_table): Remove FNNAME parameter
16847 and add SECTION parameter. Ouput one part of the table at a time.
16848 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
16849 the first part of the exception table and emit unwind directives.
16850 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
16851 (i386_pe_seh_cold_init): Likewise.
16852 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
16853 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
16854 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
16855 (ix86_output_call_insn): Emit a nop in one more case for SEH.
16856 * config/i386/winnt.c: Include except.h.
16857 (struct seh_frame_state): Add reg_offset, after_prologue and
16858 in_cold_section fields.
16859 (i386_pe_seh_end_prologue): Set seh->after_prologue.
16860 (i386_pe_seh_cold_init): New function.
16861 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
16862 to seh->in_cold_section.
16863 (seh_emit_push): Record the offset of the push.
16864 (seh_emit_save): Record the offet of the save.
16865 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
16866 Test seh->after_prologue to disregard the epilogue.
16867 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
16868 (i386_pe_end_cold_function): New function.
16870 2018-03-07 Jakub Jelinek <jakub@redhat.com>
16873 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
16874 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
16877 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
16879 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
16880 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
16882 PR middle-end/84723
16883 * multiple_target.c: Include tree-inline.h and intl.h.
16884 (expand_target_clones): Diagnose and fail if node->definition and
16885 !tree_versionable_function_p (node->decl).
16887 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
16889 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
16891 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
16892 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
16893 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
16895 2018-03-06 Jakub Jelinek <jakub@redhat.com>
16898 * combine.c (try_combine): Use reg_or_subregno instead of handling
16899 just paradoxical SUBREGs and REGs.
16901 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
16903 * config/arc/arc.c (arc_finalize_pic): Remove function.
16904 (arc_must_save_register): We use single base PIC register, remove
16905 checks to save/restore the PIC register.
16906 (arc_expand_prologue): Likewise.
16907 * config/arc/arc-protos.h (arc_set_default_type_attributes):
16909 (arc_verify_short): Likewise.
16910 (arc_attr_type): Likewise.
16911 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
16912 (walk_stores): Likewise.
16913 (arc_address_cost): Make it static.
16914 (arc_verify_short): Likewise.
16915 (branch_dest): Likewise.
16916 (arc_attr_type): Likewise.
16917 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
16918 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
16919 (arc_final_prescan_insn): Remove inserting the nops due to
16920 hardware hazards. It is done in reorg step.
16921 (insn_length_variant_t): Remove.
16922 (insn_length_parameters_t): Likewise.
16923 (arc_insn_length_parameters): Likewise.
16924 (arc_get_insn_variants): Likewise.
16925 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
16927 2018-03-06 Jakub Jelinek <jakub@redhat.com>
16929 PR inline-asm/84683
16930 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
16933 PR tree-optimization/84687
16934 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
16936 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
16938 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16940 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
16941 Rename to ppc_speculation_barrier.
16942 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
16943 __builtin_ppc_speculation_barrier.
16945 2018-03-05 Jakub Jelinek <jakub@redhat.com>
16948 * combine.c (combine_simplify_rtx): Don't try to simplify if
16949 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
16952 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
16954 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
16955 to 32 bytes when compiling for POWER9.
16957 2018-03-05 Jakub Jelinek <jakub@redhat.com>
16960 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
16961 regparm >= 3 with no arg reg available also for calls with
16962 flag_force_indirect_call. Pass decl to ix86_function_regparm.
16965 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
16967 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
16969 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
16972 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
16974 2018-03-05 Richard Biener <rguenther@suse.de>
16976 PR tree-optimization/84486
16977 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
16978 When inserting a __builtin_assume_aligned call set the LHS
16979 SSA name alignment info accordingly.
16981 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
16983 PR tree-optimization/84114
16984 * config/aarch64/aarch64.c (aarch64_reassociation_width)
16985 Avoid reassociation of FLOAT_MODE addition.
16987 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
16989 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
16990 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
16991 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
16992 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
16993 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
16994 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
16995 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
16997 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
16998 __builtin_ia32_wbinvd): New builtins.
16999 (SPECIAL_ARGS2): New.
17000 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
17001 (SPECIAL_ARGS2): New.
17002 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
17003 (ix86_valid_target_attribute_inner_p): Ditto.
17004 (ix86_init_mmx_sse_builtins): Add special_args2.
17005 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
17006 TARGET_WBNOINVD_P): New.
17007 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
17008 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
17009 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
17010 * config/i386/immintrin.h (_wbinvd): New intrinsic.
17011 * config/i386/pconfigintrin.h: New file.
17012 * config/i386/wbnoinvdintrin.h: Ditto.
17013 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
17015 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
17017 2018-03-05 Richard Biener <rguenther@suse.de>
17019 PR tree-optimization/84670
17020 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
17022 (BB_VISITED_WITH_VISITED_SUCCS): New define.
17023 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
17024 (compute_antic_aux): Only assert the number of values in ANTIC_IN
17025 doesn't grow if all successors (recursively) were visited at least
17028 2018-03-05 Richard Biener <rguenther@suse.de>
17030 PR tree-optimization/84650
17031 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
17032 if executed in the loop pipeline.
17034 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
17036 * doc/configfiles.texi (Configuration Files): Move info about
17037 conditionalizing $target-protos.h to...
17038 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
17039 differs from $target-protos.h.
17041 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
17042 Chung-Ju Wu <jasonwucj@gmail.com>
17044 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
17045 * config/nds32/nds32-multiple.md (setmemsi): Define.
17046 * config/nds32/nds32-memory-manipulation.c
17047 (nds32_gen_dup_4_byte_to_word_value): New.
17048 (emit_setmem_word_loop): New.
17049 (emit_setmem_byte_loop): New.
17050 (nds32_expand_setmem_loop): New.
17051 (nds32_expand_setmem_loop_v3m): New.
17052 (nds32_expand_setmem_unroll): New.
17053 (nds32_expand_setmem): New.
17055 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
17056 Chung-Ju Wu <jasonwucj@gmail.com>
17058 * config/nds32/nds32-memory-manipulation.c
17059 (nds32_emit_load_store): New.
17060 (nds32_emit_post_inc_load_store): New.
17061 (nds32_emit_mem_move): New.
17062 (nds32_emit_mem_move_block): New.
17063 (nds32_expand_movmemsi_loop_unknown_size): New.
17064 (nds32_expand_movmemsi_loop_known_size): New.
17065 (nds32_expand_movmemsi_loop): New.
17066 (nds32_expand_movmemsi_unroll): New.
17067 (nds32_expand_movmemqi): Rename ...
17068 (nds32_expand_movmemsi): ... to this.
17069 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
17070 (movmemsi): ... to this.
17071 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
17072 (nds32_expand_movmemsi): ... to this.
17074 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
17075 Monk Chiang <sh.chiang04@gmail.com>
17076 Chung-Ju Wu <jasonwucj@gmail.com>
17078 * config/nds32/nds32-protos.h
17079 (nds32_expand_load_multiple): New arguments.
17080 (nds32_expand_store_multiple): Ditto.
17081 (nds32_valid_multiple_load_store): Rename ...
17082 (nds32_valid_multiple_load_store_p): ... to this.
17083 * config/nds32/nds32-memory-manipulation.c
17084 (nds32_expand_load_multiple): Refine implementation.
17085 (nds32_expand_store_multiple): Ditto.
17086 * config/nds32/nds32-multiple.md
17087 (load_multiple): Update nds32_expand_load_multiple interface.
17088 (store_multiple): Update nds32_expand_store_multiple interface.
17089 * config/nds32/nds32-predicates.c
17090 (nds32_valid_multiple_load_store): Rename ...
17091 (nds32_valid_multiple_load_store_p): ... to this and refine
17093 * config/nds32/predicates.md
17094 (nds32_load_multiple_and_update_address_operation): New predicate.
17095 (nds32_store_multiple_and_update_address_operation): New predicate.
17097 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
17098 Chung-Ju Wu <jasonwucj@gmail.com>
17100 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
17101 (combo): New attribute.
17102 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
17104 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
17106 * config/nds32/nds32.opt: Change -mcmodel= default value.
17108 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
17109 Monk Chiang <sh.chiang04@gmail.com>
17110 Chung-Ju Wu <jasonwucj@gmail.com>
17112 * config/nds32/constants.md (unspec_element): New enum.
17113 * config/nds32/constraints.md (Umw): New constraint.
17114 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
17115 * config/nds32/nds32-intrinsic.md: Likewise.
17116 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
17117 (nds32_valid_smw_lwm_base_p): New.
17118 (nds32_output_smw_single_word): New.
17119 (nds32_output_lmw_single_word): New.
17120 (nds32_expand_unaligned_load): New.
17121 (nds32_expand_unaligned_store): New.
17122 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
17123 (nds32_output_smw_single_word): Declare.
17124 (nds32_output_lmw_single_word): Declare.
17125 (nds32_expand_unaligned_load): Declare.
17126 (nds32_expand_unaligned_store): Declare.
17127 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
17128 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
17129 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
17130 NDS32_BUILTIN_UASTORE_DW.
17131 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
17134 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
17135 Kito Cheng <kito.cheng@gmail.com>
17136 Chung-Ju Wu <jasonwucj@gmail.com>
17138 * config/nds32/nds32-intrinsic.c
17139 (nds32_expand_builtin_null_ftype_reg): Delete.
17140 (nds32_expand_builtin_reg_ftype_imm): Ditto.
17141 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
17142 (nds32_read_argument): New.
17143 (nds32_legitimize_target): Ditto.
17144 (nds32_legitimize_argument): Ditto.
17145 (nds32_check_constant_argument): Ditto.
17146 (nds32_expand_unop_builtin): Ditto.
17147 (nds32_expand_unopimm_builtin): Ditto.
17148 (nds32_expand_binop_builtin): Ditto.
17149 (nds32_builtin_decl_impl): Ditto.
17150 (builtin_description): Ditto.
17151 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
17152 (nds32_init_builtins_impl): Ditto.
17153 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
17154 (nds32_builtin_decl): New.
17155 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
17156 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
17158 2018-03-02 Jeff Law <law@redhat.com>
17160 * reorg.c (stop_search_p): Handle DEBUG_INSN.
17161 (redundant_insn, fill_simple_delay_slots): Likewise.
17162 (fill_slots_from_thread): Likewise.
17163 * resource.c (mark_referenced_resources): Likewise.
17164 (mark_set_resources, find_dead_or_set_registers): Likewise.
17166 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17168 * substring-locations.h (format_warning_va): Formatting fix for
17169 ATTRIBUTE_GCC_DIAG.
17170 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
17172 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
17173 * substring-locations.c: Include intl.h.
17174 (format_warning_va): Turned into small wrapper around
17175 format_warning_n_va, renamed to ...
17176 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
17177 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
17179 (format_warning_at_substring_n): New function.
17180 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
17181 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
17182 format_warning_at_substring with just a shorter name instead of
17183 const function pointer.
17184 (fmtwarn_n): New function.
17185 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
17186 appropriate, get rid of all the fmtstr temporaries, move conditionals
17187 with G_() wrapped string literals directly into fmtwarn arguments,
17188 cast dir.len to (int), formatting fixes.
17190 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
17192 * doc/invoke.texi: Remove "Cilk Plus" references.
17194 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17195 Richard Biener <rguenther@suse.de>
17198 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
17199 for error or warning attributes if CALL_FROM_THUNK_P is set.
17202 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17205 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
17206 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
17209 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
17210 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
17212 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
17213 instead of -1U in last predictors element's probability member.
17215 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
17218 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
17219 arguments if they are comparable.
17221 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
17223 PR tree-optimization/84634
17224 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
17225 masks and masked_loop_p with a single loop_masks, making sure it's
17226 null for bb vectorization.
17228 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
17230 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
17231 (vect_analyze_data_ref_access): Use loop->safe_len rather than
17232 loop->force_vectorize to check whether there is no alias.
17234 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17237 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
17239 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
17241 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
17242 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
17243 instead of a loop around prev_real_insn.
17244 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
17247 PR inline-asm/84625
17248 * config/i386/i386.c (ix86_print_operand): Use conditional
17249 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
17252 2018-03-02 Richard Biener <rguenther@suse.de>
17254 PR tree-optimization/84427
17255 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
17256 (bitmap_set_subtract_values): Rewrite to handle multiple
17259 (prune_clobbered_mems): Likewise.
17260 (phi_translate): Take edge instead of pred/phiblock.
17261 (phi_translate_1): Likewise.
17262 (phi_translate_set): Likewise. Insert all translated
17263 exprs for a value into the set, keeping possibly multiple
17264 expressions per value.
17265 (compute_antic_aux): Adjust for phi_translate changes.
17266 When intersecting union the expressions and prune those
17267 not in the final value set, keeping possibly multiple
17268 expressions per value. Do not use value-insertion
17269 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
17270 all expressions. Add verification that the value-sets
17271 only shrink during iteration.
17272 (compute_partial_antic_aux): Adjust for the phi_translate changes.
17273 (do_pre_regular_insertion): Likewise.
17274 (do_pre_partial_partial_insertion): Likewise.
17276 2018-03-02 Richard Biener <rguenther@suse.de>
17279 * config/darwin.c (saved_debug_info_level): New static global.
17280 (darwin_asm_lto_start): Disable debug info generation for LTO out.
17281 (darwin_asm_lto_end): Restore debug info generation settings.
17283 2018-03-01 Martin Liska <mliska@suse.cz>
17286 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
17287 volatile arguments.
17289 2018-03-01 Richard Biener <rguenther@suse.de>
17292 * dwarf2out.c (gen_variable_die): Properly handle late VLA
17293 type annotation with LTO when debug was disabled at compile-time.
17295 2018-03-01 Matthew Fortune <mfortune@gmail.com>
17297 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
17299 (mips_final_postscan_insn): Likewise.
17301 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
17303 PR rtl-optimization/84528
17304 * alias.c (init_alias_target): Add commentary.
17305 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
17306 a unique base value if the frame pointer is not eliminated
17307 to the stack pointer.
17309 2018-03-01 Tom de Vries <tom@codesourcery.com>
17311 PR rtl-optimization/83327
17312 * lra-int.h (hard_regs_spilled_into): Declare.
17313 * lra.c (hard_regs_spilled_into): Define.
17314 (init_reg_info): Init hard_regs_spilled_into.
17315 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
17316 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
17317 (process_bb_lives): Handle hard_regs_spilled_into.
17318 (lra_create_live_ranges_1): Before doing liveness propagation, clear
17319 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
17321 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
17323 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
17324 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
17325 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
17326 * config/rs6000/aix72.h: New file.
17328 2018-02-28 Jakub Jelinek <jakub@redhat.com>
17330 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
17331 instead of warning_at with conditional singular and plural messages
17335 * stor-layout.c (update_alignment_for_field): For
17336 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
17337 && !DECL_PACKED (field), do the alignment update, just use
17338 only desired_align instead of MAX (type_align, desired_align)
17340 (place_field): Don't do known_align < desired_align handling
17341 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
17342 is non-NULL, instead do it after rli->prev_field handling and
17343 only if not within a bitfield word. For DECL_PACKED (field)
17344 use type_align of BITS_PER_UNIT.
17346 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
17348 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
17349 superfluous parentheses and trailing spaces.
17351 2018-02-28 Richard Biener <rguenther@suse.de>
17353 PR tree-optimization/84584
17354 * graphite-scop-detection.c (scop_detection::add_scop): Discard
17355 SCoPs with fake exit edge.
17357 2018-02-28 Martin Liska <mliska@suse.cz>
17360 * timevar.c (timer::print): Fix format to properly print 100%
17363 2018-02-28 Richard Biener <rguenther@suse.de>
17365 PR middle-end/84607
17366 * genmatch.c (capture_info::walk_match): Do not mark
17367 captured expressions without operands as expr_p given
17368 they act more like predicates and should be subject to
17369 "lost tail" side-effect preserving.
17371 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
17373 PR rtl-optimization/81611
17374 * auto-inc-dec.c (attempt_change): Move dead note from
17375 mem_insn if it's the next use of regno
17376 (find_address): Take address use of reg holding
17377 non-incremented value. Add parm to limit search to the named
17379 (merge_in_block): Attempt to use a mem insn that is the next
17380 use of the original regno.
17382 2018-02-27 Martin Sebor <msebor@redhat.com>
17385 * doc/invoke.texi (-Wmissing-attributes): New option.
17386 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
17388 2018-02-27 Martin Sebor <msebor@redhat.com>
17390 PR translation/84207
17391 * diagnostic-core.h (warning_n, error_n, inform_n): Change
17392 n argument to unsigned HOST_WIDE_INT.
17393 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
17394 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
17395 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
17396 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
17398 2018-02-27 Richard Biener <rguenther@suse.de>
17400 PR tree-optimization/84512
17401 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
17402 Do not use the estimate returned from record_stmt_cost for
17403 the scalar iteration cost but sum properly using add_stmt_cost.
17405 2018-02-27 Richard Biener <rguenther@suse.de>
17407 PR tree-optimization/84466
17408 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
17409 Adjust last change to less strictly validate use operands.
17411 2018-02-27 Martin Liska <mliska@suse.cz>
17413 PR gcov-profile/84548
17414 * gcov.c (process_file): Allow partial overlap and consider it
17415 also as group functions.
17416 (output_lines): Properly calculate range of lines for a group.
17418 2018-02-27 Martin Liska <mliska@suse.cz>
17420 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
17421 'ggc' suffixes. Change first column width.
17422 (timer::print): Fix formatting of the column.
17424 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
17426 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
17427 preserve inline entry blocks for the sake of debug inline
17428 entry point markers alone.
17429 (remove_unused_locals): Suggest in comments a better place to
17430 force the preservation of inline entry blocks that are
17431 otherwise unused, but do not preserve them.
17433 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
17435 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
17437 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
17440 * config/i386/constraints.md (Bs): Replace
17441 ix86_indirect_branch_register with
17442 TARGET_INDIRECT_BRANCH_REGISTER.
17444 * config/i386/i386.md (indirect_jump): Likewise.
17445 (tablejump): Likewise.
17446 (*sibcall_memory): Likewise.
17447 (*sibcall_value_memory): Likewise.
17448 Peepholes of indirect call and jump via memory: Likewise.
17449 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
17450 (*sibcall_value_GOT_32): Likewise.
17451 * config/i386/predicates.md (indirect_branch_operand): Likewise.
17452 (GOT_memory_operand): Likewise.
17453 (call_insn_operand): Likewise.
17454 (sibcall_insn_operand): Likewise.
17455 (GOT32_symbol_operand): Likewise.
17456 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
17458 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
17460 PR rtl-optimization/83496
17461 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
17462 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
17463 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
17464 redundant insn, if any.
17465 (relax_delay_slots): Likewise.
17466 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
17468 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
17470 PR tree-optimization/83965
17471 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
17472 that grouped statements are part of a reduction chain. Return
17473 true if the statement is not marked as a reduction itself but
17474 is part of a group.
17475 (vect_recog_dot_prod_pattern): Don't check whether the statement
17476 is part of a group here.
17477 (vect_recog_sad_pattern): Likewise.
17478 (vect_recog_widen_sum_pattern): Likewise.
17480 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
17483 * final.c (rest_of_clean_state): Also look for calls inside sequences.
17485 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
17488 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
17490 (ix86_output_indirect_function_return): New prototype.
17491 (ix86_split_simple_return_pop_internal): Likewise.
17492 * config/i386/i386.c (indirect_return_via_cx): New.
17493 (indirect_return_via_cx_bnd): Likewise.
17494 (indirect_thunk_name): Handle return va CX_REG.
17495 (output_indirect_thunk_function): Create alias for
17496 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
17497 (ix86_output_indirect_jmp): Remove the bool argument.
17498 (ix86_output_indirect_function_return): New function.
17499 (ix86_split_simple_return_pop_internal): Likewise.
17500 * config/i386/i386.md (*indirect_jump): Don't pass false
17501 to ix86_output_indirect_jmp.
17502 (*tablejump_1): Likewise.
17503 (simple_return_pop_internal): Change it to define_insn_and_split.
17504 Call ix86_split_simple_return_pop_internal to split it for
17505 -mfunction-return=.
17506 (simple_return_indirect_internal): Call
17507 ix86_output_indirect_function_return instead of
17508 ix86_output_indirect_jmp.
17510 2018-02-26 Jakub Jelinek <jakub@redhat.com>
17513 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
17514 memset and value initialization afterwards.
17516 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
17518 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
17520 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
17523 * common/config/aarch64/aarch64-common.c
17524 (aarch_option_optimization_table[]): Switch
17525 off fomit-frame-pointer
17527 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
17528 Chung-Ju Wu <jasonwucj@gmail.com>
17530 * config/nds32/nds32-multiple.md (load_multiple): Disallow
17532 (store_multiple): Ditto.
17534 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
17536 * config.gcc: Add --with-cpu support for nds32 target.
17537 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
17538 * config/nds32/nds32.opt: Add -mcpu= option.
17540 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
17542 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
17543 isel=yes): Warn for these deprecated options.
17545 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
17547 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
17548 ISA_2_5_MASKS_EMBEDDED.
17550 2018-02-23 Jakub Jelinek <jakub@redhat.com>
17552 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
17553 p->max as pointers rather than using iterative_hash_expr.
17555 2018-02-23 Carl Love <cel@us.ibm.com>
17557 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
17558 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
17560 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
17561 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
17562 P8V_BUILTIN_VEC_VUNSIGNED2.
17564 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
17567 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
17568 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
17570 * lra-constraints.c (curr_insn_transform): Set up
17571 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
17573 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
17574 * lra-eliminations.c (spill_pseudos): Ditto.
17575 (process_insn_for_elimination): Ditto.
17576 * lra-lives.c (reg_early_clobber_p): Use the new macros.
17577 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
17578 LRA_NON_CLOBBERED_ALT.
17580 2018-02-22 Martin Sebor <msebor@redhat.com>
17582 PR tree-optimization/84480
17583 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
17584 to maybe_diag_stxncpy_trunc. Call it.
17585 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
17586 from gimple_fold_builtin_strcpy. Print inlining stack.
17587 (handle_builtin_stxncpy): Print inlining stack.
17588 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
17590 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
17593 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
17594 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
17595 and -fcheck-pointer-bounds are used together.
17596 (indirect_thunk_prefix): New enum.
17597 (indirect_thunk_need_prefix): New function.
17598 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
17599 "_nt" instead of "_bnd" for NOTRACK prefix.
17600 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
17601 (output_indirect_thunk_function): Likewise.
17603 (ix86_code_end): Update output_indirect_thunk_function calls.
17604 (ix86_output_indirect_branch_via_reg): Replace
17605 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
17606 (ix86_output_indirect_branch_via_push): Likewise.
17607 (ix86_output_function_return): Likewise.
17608 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
17609 incompatible with -fcf-protection=branch and
17610 -fcheck-pointer-bounds.
17612 2018-02-22 Steve Ellcey <sellcey@cavium.com>
17615 * config/aarch64/aarch64.c (aarch64_print_address_internal):
17616 Change gcc_assert call to output_operand_lossage.
17618 2018-02-22 Steve Ellcey <sellcey@cavium.com>
17620 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
17622 2018-02-22 DJ Delorie <dj@redhat.com>
17623 Sebastian Perta <sebastian.perta@renesas.com>
17624 Oleg Endo <olegendo@gcc.gnu.org>
17626 * config/rx/rx.c (rx_rtx_costs): New function.
17627 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
17629 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
17631 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
17633 2018-02-22 Martin Liska <mliska@suse.cz>
17636 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
17637 Add "native" as a possible value.
17639 2018-02-22 Martin Liska <mliska@suse.cz>
17642 * config/i386/i386.c (ix86_option_override_internal):
17643 Add "native" as a possible value for -march and -mtune.
17645 2018-02-22 Jakub Jelinek <jakub@redhat.com>
17648 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
17649 to all type variants.
17651 PR tree-optimization/84503
17652 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
17653 width as info->bitpos + info->bitsize - start.
17654 (merged_store_group::merge_overlapping): Simplify width computation.
17655 (check_no_overlap): New function.
17656 (imm_store_chain_info::try_coalesce_bswap): Compute expected
17657 start + width and last_order of the group, fail if check_no_overlap
17659 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
17660 to group if check_no_overlap fails.
17662 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
17664 * config/rs6000/altivec.md: Delete contraint arguments to
17665 define_expand, define_split, and define_peephole2, and in
17666 define_insn_and_split if always unused.
17667 * config/rs6000/darwin.md: Ditto.
17668 * config/rs6000/dfp.md: Ditto.
17669 * config/rs6000/rs6000.md: Ditto.
17670 * config/rs6000/sync.md: Ditto.
17671 * config/rs6000/vector.md: Ditto.
17672 * config/rs6000/vsx.md: Ditto.
17674 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
17676 * config/rs6000/altivec.md: Write output control strings as braced
17677 blocks instead of double-quoted strings.
17678 * config/rs6000/darwin.md: Ditto.
17679 * config/rs6000/rs6000.md: Ditto.
17680 * config/rs6000/vector.md: Ditto.
17681 * config/rs6000/vsx.md: Ditto.
17683 2018-02-21 Jason Merrill <jason@redhat.com>
17685 PR c++/84314 - ICE with templates and fastcall attribute.
17686 * attribs.c (build_type_attribute_qual_variant): Remove assert.
17688 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
17690 * ipa-cp.c (determine_versionability): Fix comment typos.
17692 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
17695 * ipa-cp.c (determine_versionability): Do not version functions caling
17698 2018-02-21 Martin Liska <mliska@suse.cz>
17701 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
17702 Add "native" as a possible value.
17703 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
17704 the macro when native cpu detection is available.
17706 2018-02-21 Martin Liska <mliska@suse.cz>
17709 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
17710 Add "native" as a possible value.
17711 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
17712 when native cpu detection is available.
17714 2018-02-21 Jakub Jelinek <jakub@redhat.com>
17715 Martin Sebor <msebor@redhat.com>
17717 PR tree-optimization/84478
17718 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
17720 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
17721 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
17722 support which is conservatively correct, for 2 only stay conservative
17723 for maxlen. Formatting and comment capitalization fixes. Add STRICT
17724 argument to the 2 argument get_range_strlen, adjust 6 arg
17725 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
17727 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
17728 (gimple_fold_builtin_strlen): Pass true as last argument to
17731 2018-02-20 Martin Sebor <msebor@redhat.com>
17733 PR middle-end/84095
17734 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
17735 (builtin_memref::set_base_and_offset): Same. Handle inner references.
17736 (builtin_memref::builtin_memref): Factor out parts into
17737 set_base_and_offset and call it.
17739 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
17741 PR middle-end/84406
17742 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
17743 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
17744 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
17745 search at the associated MODE_INT.
17747 2018-02-20 Jeff Law <law@redhat.com>
17749 PR middle-end/82123
17750 PR tree-optimization/81592
17751 PR middle-end/79257
17752 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
17753 for range data rather than using global data.
17754 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
17755 range data rather than using global data.
17756 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
17757 pass it to children as needed.
17758 (struct directive::fmtresult): Similarly.
17759 (struct directive::set_width): Similarly.
17760 (struct directive::set_precision): Similarly.
17761 (format_integer, format_directive, parse_directive): Similarly.
17762 (format_none): Accept unnamed vr_values parameter.
17763 (format_percent, format_floating, format_character): Similarly.
17764 (format_string, format_plain): Similarly.
17765 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
17766 the EVRP range analyzer for range data rather than using global data.
17767 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
17768 gimple-ssa-evrp-analyze.h
17769 (class sprintf_dom_walker): Add after_dom_children member function.
17770 Add evrp_range_analyzer member.
17771 (sprintf_dom_walker::before_dom_children): Call into the EVRP
17772 range analyzer as needed.
17773 (sprintf_dom_walker::after_dom_children): New member function.
17774 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
17776 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
17777 (evrp_range_analyzer::pop_to_marker): Likewise.
17779 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
17781 PR tree-optimization/84419
17782 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
17783 with the required type if its current type is compatible but
17786 2018-02-20 Jakub Jelinek <jakub@redhat.com>
17788 PR middle-end/82004
17789 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
17790 after vectorization.
17792 2018-02-20 Martin Liska <mliska@suse.cz>
17795 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
17796 possible values if we don't have a hint.
17798 2018-02-20 Martin Liska <mliska@suse.cz>
17802 * final.c (shorten_branches): Build align_tab array with one
17804 * opts.c (finish_options): Add alignment option limit check.
17805 (MAX_CODE_ALIGN): Likewise.
17806 (MAX_CODE_ALIGN_VALUE): Likewise.
17807 * doc/invoke.texi: Document maximum allowed option value for
17808 all -falign-* options.
17810 2018-02-19 Jakub Jelinek <jakub@redhat.com>
17813 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
17814 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
17815 * var-tracking.c (emit_note_insn_var_location): Remove all references
17816 to NOTE_INSN_CALL_ARG_LOCATION.
17817 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
17818 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
17819 Use copy_rtx_if_shared.
17820 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
17821 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
17822 (dwarf2out_var_location): Remove handling of
17823 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
17825 * final.c (final_scan_insn): Remove all references to
17826 NOTE_INSN_CALL_ARG_LOCATION.
17827 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
17828 before dumping final insns.
17829 * except.c (emit_note_eh_region_end): Remove all references to
17830 NOTE_INSN_CALL_ARG_LOCATION.
17831 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
17832 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
17833 * config/arc/arc.c (hwloop_optimize): Likewise.
17834 * config/arm/arm.c (create_fix_barrier): Likewise.
17835 * config/s390/s390.c (s390_chunkify_start): Likewise.
17836 * config/sh/sh.c (find_barrier): Likewise.
17837 * config/i386/i386.c (rest_of_insert_endbranch,
17838 ix86_seh_fixup_eh_fallthru): Likewise.
17839 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
17840 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
17841 * config/frv/frv.c (frv_function_prologue): Likewise.
17842 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
17844 (note_outside_basic_block_p): Remove all references to
17845 NOTE_INSN_CALL_ARG_LOCATION.
17846 * gengtype.c (adjust_field_rtx_def): Likewise.
17847 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
17849 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
17850 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
17853 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
17856 PR tree-optimization/84452
17857 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
17858 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
17861 2018-02-19 Martin Liska <mliska@suse.cz>
17864 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
17866 2018-02-19 Martin Liska <mliska@suse.cz>
17867 Richard Sandiford <richard.sandiford@linaro.org>
17869 PR tree-optimization/82491
17870 * gimple-fold.c (get_base_constructor): Make earlier bail out
17873 2018-02-19 Carl Love <cel@us.ibm.com>
17875 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
17876 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
17878 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
17879 P8V_BUILTIN_VEC_NEG.
17881 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
17883 * config/rl78/rl78.md (movdf): New define expand.
17885 2018-02-19 Martin Liska <mliska@suse.cz>
17888 * doc/invoke.texi: Fix typo.
17889 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
17891 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
17893 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
17894 handle rs6000_single_float and rs6000_double_float specially for
17897 2018-02-16 Jeff Law <law@redhat.com>
17899 * config/rx/rx.c (add_pop_cfi_notes): New function.;
17900 (pop_regs): Use it.
17902 2018-02-16 Jakub Jelinek <jakub@redhat.com>
17905 * ipa-inline.c (inline_small_functions): Fix a typo.
17907 2018-02-16 Nathan Sidwell <nathan@acm.org>
17909 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
17911 2018-02-16 Carl Love <cel@us.ibm.com>
17913 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
17914 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
17915 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
17916 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
17917 expansion to P8V_BUILTIN_VEC_FLOAT2.
17919 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
17921 PR rtl-optimization/70023
17922 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
17923 src_regno into account.
17925 2018-02-16 Carl Love <cel@us.ibm.com>
17927 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
17928 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
17929 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
17930 * config/rs6000/rs6000.c: Remove case statements for
17931 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
17932 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
17933 and P9V_BUILTIN_VEC_VINSERT4B.
17934 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
17935 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
17936 * config/rs6000/vsx.md:
17937 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
17940 2018-02-16 Carl Love <cel@us.ibm.com>
17942 * config/rs6000/altivec.h: Add builtin names vec_extract4b
17944 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
17946 * config/rs6000/rs6000-c.c: Add the definitions for
17947 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
17948 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
17949 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
17950 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
17951 definition for insert4b and define insn *insert3b_internal.
17952 * doc/extend.texi: Add documentation for vec_extract4b.
17954 2018-02-16 Nathan Sidwell <nathan@acm.org>
17956 * doc/extend.texi (Backwards Compatibility): Mention friend
17957 injection. Note for-scope is deprecated.
17958 * doc/invoke.texi (-ffriend-injection): Deprecate.
17960 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
17962 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
17963 that moved to I2, also allow destinations that are a paradoxical
17964 subreg (instead of a normal reg).
17966 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
17969 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
17972 2018-02-16 Richard Biener <rguenther@suse.de>
17974 PR tree-optimization/84037
17975 PR tree-optimization/84016
17977 * config/i386/i386.c (ix86_builtin_vectorization_cost):
17978 Adjust vec_construct for the fact we need additional higher latency
17979 128bit inserts for AVX256 and AVX512 vector builds.
17980 (ix86_add_stmt_cost): Scale vector construction cost for
17983 2018-02-16 Richard Biener <rguenther@suse.de>
17985 PR tree-optimization/84417
17986 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
17987 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
17988 (non_rewritable_lvalue_p): Likewise, use poly-ints.
17990 2018-02-16 Martin Liska <mliska@suse.cz>
17993 * internal-fn.def (ASAN_CHECK): Set proper flags.
17994 (ASAN_MARK): Likewise.
17996 2018-02-16 Julia Koval <julia.koval@intel.com>
17998 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
17999 from PTA_CANNONLAKE.
18001 2018-02-16 Jakub Jelinek <jakub@redhat.com>
18004 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
18005 Use ++iter rather than iter++ for std::list iterators.
18006 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
18007 defer deleting them until all nodes in the forest are processed. Do
18008 free even leaf nodes. Change to_process into auto_vec.
18011 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
18012 * vec.h (vec_default_construct): Use memset instead of placement new
18013 if BROKEN_VALUE_INITIALIZATION is defined.
18014 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
18015 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
18018 PR rtl-optimization/83723
18019 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
18020 * lra.c (lra_substitute_pseudo): Likewise. If true, use
18021 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
18023 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
18025 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
18027 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
18029 PR rtl-optimization/81443
18030 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
18031 from inner REGs to paradoxical SUBREGs.
18033 2018-02-16 Richard Biener <rguenther@suse.de>
18035 PR tree-optimization/84399
18036 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
18037 For operands we can analyze at their definition make sure we can
18038 analyze them at each use as well.
18040 2018-02-16 Richard Biener <rguenther@suse.de>
18042 PR tree-optimization/84190
18043 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
18044 volatile accesses if the decl isn't volatile.
18046 2018-02-15 Jason Merrill <jason@redhat.com>
18048 PR c++/84314 - ICE with templates and fastcall attribute.
18049 * attribs.c (build_type_attribute_qual_variant): Don't clobber
18050 TYPE_CANONICAL on an existing type.
18052 2018-02-15 Jakub Jelinek <jakub@redhat.com>
18054 PR tree-optimization/84383
18055 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
18056 dstoff nor call operand_equal_p if dstbase is NULL.
18058 PR tree-optimization/84334
18059 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
18060 also a CONSTANT_CLASS_P, punt.
18062 2018-02-14 Jim Wilson <jimw@sifive.com>
18064 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
18065 first SMALL_OPERAND check. New local min_second_step. Move assert
18066 to where locals are set. Add TARGET_RVC support.
18067 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
18069 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
18071 * doc/invoke.texi: Correct -Wformat-overflow code sample.
18073 2018-02-14 Martin Sebor <msebor@redhat.com>
18075 PR tree-optimization/83698
18076 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
18077 arrays constrain the offset range to their bounds.
18078 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
18079 (builtin_access::overlap): Avoid setting the size of overlap if it's
18081 (maybe_diag_overlap): Also consider arrays when deciding what values
18082 of offsets to include in diagnostics.
18084 2018-02-14 Martin Sebor <msebor@redhat.com>
18087 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
18088 that correspond to the kind of a declaration.
18090 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
18093 * config/pa/pa.md: Load address of PIC label using the linkage table
18094 if the label is nonlocal.
18096 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
18098 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
18099 warning message if user requests -maltivec=be.
18100 * doc/invoke.texi: Document deprecation of -maltivec=be.
18102 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
18105 * config/rs6000/rs6000-c.c: Update definitions for
18106 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
18107 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
18109 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
18112 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
18113 add _get_ssp intrinsics. Remove argument from
18114 __builtin_ia32_rdssp[d|q].
18115 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
18116 * config/i386/i386-builtin.def: Remove argument from
18117 __builtin_ia32_rdssp[d|q].
18118 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
18119 ix86_expand_special_args_builtin for _rdssp[d|q].
18120 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
18121 Clear register before usage.
18122 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
18123 Add documentation for new _get_ssp and _inc_ssp intrinsics.
18125 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
18127 PR tree-optimization/84357
18128 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
18129 operand 1 of an ARRAY_REF too.
18131 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
18134 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
18135 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
18137 (set_of_reg): New struct.
18138 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
18139 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
18140 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
18142 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
18143 Split into bitclr, bitset, bitinvert patterns if appropriate.
18144 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
18145 use rx_fuse_in_memory_bitop.
18146 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
18147 to named insn, correct maximum insn length.
18149 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
18152 * machmode.def: Define a complex mode for PARTIAL_INT.
18153 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
18155 * doc/rtl.texi: Document CSPImode.
18156 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
18158 (msp430_hard_regno_nregs_with_padding): Likewise.
18160 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
18163 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
18165 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
18167 PR rtl-optimization/84169
18168 * combine.c (try_combine): New variable split_i2i3. Set it to true if
18169 we generated a parallel as new i3 and we split that to new i2 and i3
18170 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
18171 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
18172 those to i2, not i1. Partially rewrite this scan code.
18174 2018-02-13 Jakub Jelinek <jakub@redhat.com>
18177 * stor-layout.c (place_field): For variable length fields, adjust
18178 offset_align afterwards not just based on the field's alignment,
18179 but also on the size.
18181 PR middle-end/84309
18182 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
18183 of exps and logs in the use_exp2 case.
18185 2018-02-13 Jeff Law <law@redhat.com>
18187 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
18188 entry for "vector".
18190 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
18193 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
18197 * common.opt (gas-loc-support, gas-locview-support): New.
18198 (ginline-points, ginternal-reset-location-views): New.
18199 * doc/invoke.texi: Document them. Use @itemx where intended.
18200 (gvariable-location-views): Adjust.
18201 * target.def (reset_location_view): New.
18202 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
18203 (TARGET_RESET_LOCATION_VIEW): New.
18204 * doc/tm.texi: Rebuilt.
18205 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
18206 (dwarf2out_default_as_locview_support): New.
18207 (output_asm_line_debug_info): Use option variables.
18208 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
18209 (output_loc_list): Likewise.
18210 (add_high_low_attributes): Check option variables.
18211 Don't output entry view attribute in strict mode.
18212 (gen_inlined_subroutine_die): Check option variables.
18213 (dwarf2out_inline_entry): Likewise.
18214 (init_sections_and_labels): Likewise.
18215 (dwarf2out_early_finish): Likewise.
18216 (maybe_reset_location_view): New, from...
18217 (dwarf2out_var_location): ... here. Call it.
18218 * debug.h (dwarf2out_default_as_loc_support): Declare.
18219 (dwarf2out_default_as_locview_support): Declare.
18220 * hooks.c (hook_int_rtx_insn_0): New.
18221 * hooks.h (hook_int_rtx_insn_0): Declare.
18222 * toplev.c (process_options): Take -gas-loc-support and
18223 -gas-locview-support from dwarf2out. Enable
18224 -gvariable-location-views by default only with locview
18225 assembler support. Enable -ginternal-reset-location-views by
18226 default only if the target defines the corresponding hook.
18227 Enable -ginline-points by default if location views are
18228 enabled; force it disabled if statement frontiers are
18230 * tree-inline.c (expand_call_inline): Check option variables.
18231 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
18233 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
18235 PR tree-optimization/84321
18236 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
18237 handling. Also check whether the anti-range contains any values
18238 that satisfy the mask; switch to a VR_RANGE if not.
18240 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
18243 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
18245 2018-02-13 Martin Jambor <mjambor@suse.cz>
18248 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
18249 of call statements, also set location of a load to a temporary.
18251 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
18253 * config/rl78/rl78.c (add_vector_labels): New function.
18254 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
18255 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
18256 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
18257 which checks that no arguments are passed.
18258 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
18259 * doc/extend.texi: Documentation for the new attribute.
18261 2018-02-13 Andreas Schwab <schwab@suse.de>
18263 * config/riscv/linux.h (CPP_SPEC): Define.
18265 2018-02-13 Jakub Jelinek <jakub@redhat.com>
18268 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
18269 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
18270 OPTION_MASK_ISA_AES as first argument to def_builtin_const
18271 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
18272 instead of OPTION_MASK_ISA_PCLMUL as first argument to
18273 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
18274 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
18275 temporarily for AES and PCLMUL builtins.
18277 PR tree-optimization/84339
18278 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
18279 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
18282 PR middle-end/84309
18283 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
18284 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
18285 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
18287 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
18289 * omp-simd-clone.h: New file.
18290 * omp-simd-clone.c: Include omp-simd-clone.h.
18291 (expand_simd_clones): No longer static.
18292 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
18293 cgraph.h and omp-simd-clone.h.
18294 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
18295 (vect_recog_widen_shift_pattern): Formatting fix.
18296 (vect_pattern_recog_1): Don't check optab for calls.
18299 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
18300 operands[2] into a REG before using gen_lowpart on it.
18302 2018-02-12 Jeff Law <law@redhat.com>
18305 * config/sh/sh.c (find_barrier): Consider a sibling call
18308 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
18309 successfully back substituting a reg.
18311 2018-02-12 Richard Biener <rguenther@suse.de>
18313 PR tree-optimization/84037
18314 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
18315 parameter, move visited init to caller.
18316 (vect_slp_analyze_operations): Separate cost from validity
18317 check, initialize visited once for all instances.
18318 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
18320 * tree-vect-stmts.c (vect_model_simple_cost): Make early
18322 (vect_model_promotion_demotion_cost): Likewise.
18323 (vectorizable_bswap): Guard cost modeling with !slp_node
18324 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
18326 (vectorizable_call): Likewise.
18327 (vectorizable_conversion): Likewise.
18328 (vectorizable_assignment): Likewise.
18329 (vectorizable_shift): Likewise.
18330 (vectorizable_operation): Likewise.
18331 (vectorizable_store): Likewise.
18332 (vectorizable_load): Likewise.
18333 (vectorizable_condition): Likewise.
18334 (vectorizable_comparison): Likewise.
18336 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
18339 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
18340 (ASAN_MARK): Fix fnspec to account for return value, change pointer
18341 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
18343 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
18345 PR middle-end/83665
18346 * params.def (inline-min-speedup): Increase from 8 to 15.
18347 (max-inline-insns-auto): Decrease from 40 to 30.
18348 * ipa-split.c (consider_split): Add some buffer for function to
18349 be considered inlining candidate.
18350 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
18353 2018-02-12 Richard Biener <rguenther@suse.de>
18355 PR tree-optimization/84037
18356 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
18357 matched stmts if we cannot swap the non-matched ones.
18359 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
18361 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
18362 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
18363 _mm_maskz_scalef_round_ss): New intrinsics.
18364 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
18365 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
18366 __builtin_ia32_scalefss_round): Remove.
18367 (__builtin_ia32_scalefsd_mask_round,
18368 __builtin_ia32_scalefss_mask_round): New intrinsics.
18369 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
18370 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
18371 ((match_operand:VF_128 2 "<round_nimm_predicate>"
18372 "<round_constraint>")): Changed to ...
18373 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
18374 "<round_scalar_constraint>")): ... this.
18375 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
18376 %0, %1, %2<round_op3>}"): Changed to ...
18377 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
18378 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
18379 %2<round_scalar_mask_op3>}"): ... this.
18380 * config/i386/subst.md (round_scalar_nimm_predicate): New.
18382 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
18384 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
18385 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
18386 (_mm_maskz_sqrt_round_ss): New intrinsics.
18387 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
18388 (__builtin_ia32_sqrtsd_mask_round)
18389 (__builtin_ia32_sqrtss_mask_round): New builtins.
18390 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
18391 (__builtin_ia32_sqrtss_round): Remove.
18392 (__builtin_ia32_sqrtsd_mask_round)
18393 (__builtin_ia32_sqrtss_mask_round): New builtins.
18394 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
18395 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
18396 ((match_operand:VF_128 1 "vector_operand"
18397 "xBm,<round_constraint>")): Changed to ...
18398 ((match_operand:VF_128 1 "vector_operand"
18399 "xBm,<round_scalar_constraint>")): ... this.
18400 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
18401 %0, %2, %<iptr>1<round_op3>}): Changed to ...
18402 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
18403 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
18404 %<iptr>1<round_scalar_mask_op3>}): ... this.
18405 ((set_attr "prefix" "<round_prefix>")): Changed to ...
18406 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
18408 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
18411 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
18412 Cast vec_cmpeq result to correct type.
18413 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
18414 Cast vec_cmpgt result to correct type.
18416 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
18418 * final.c (final_scan_insn_1): Renamed from...
18419 (final_scan_insn): ... this. New wrapper, to recover
18420 seen from the outermost call in recursive ones.
18421 * config/sparc/sparc.c (output_return): Drop seen from call.
18422 (output_sibcall): Likewise.
18423 * config/visium/visium.c (output_branch): Likewise.
18425 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
18427 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
18430 2018-02-10 Alan Modra <amodra@gmail.com>
18433 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
18434 Specify LR as an input.
18436 2018-02-10 Jakub Jelinek <jakub@redhat.com>
18439 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
18440 remove_member_access_dummy_vars): New functions.
18441 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
18442 lower_omp_1, execute_lower_omp): Use them.
18444 PR rtl-optimization/84308
18445 * shrink-wrap.c (spread_components): Release todo vector.
18447 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
18449 PR rtl-optimization/57193
18450 * ira-color.c (struct allocno_color_data): Add member
18451 conflict_allocno_hard_prefs.
18452 (update_conflict_allocno_hard_prefs): New.
18453 (bucket_allocno_compare_func): Add a preference based on
18454 conflict_allocno_hard_prefs.
18455 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
18456 (color_allocnos): Remove a dead code. Initiate
18457 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
18459 2018-02-09 Jakub Jelinek <jakub@redhat.com>
18462 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
18463 constraint from =wa to wa. Avoid a subreg on the output operand,
18464 instead use a pseudo and subreg it in a move.
18465 (p9_xxbrd_<mode>): Changed to ...
18466 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
18467 (p9_xxbrd_v2df): New expander.
18468 (p9_xxbrw_<mode>): Changed to ...
18469 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
18470 (p9_xxbrw_v4sf): New expander.
18472 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
18474 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
18476 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
18479 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
18480 multiply in 32-bit mode.
18481 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
18482 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
18485 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
18487 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
18488 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
18489 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
18490 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
18492 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
18495 * dwarf2out.c (is_trivial_indirect_ref): New function.
18496 (dwarf2out_late_global_decl): Do not generate a location
18497 attribute for variables that have a non-trivial DECL_VALUE_EXPR
18498 and that are not defined in the current unit.
18500 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
18502 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
18503 instead of a libcall for UNORDERED.
18505 2018-02-09 Tamar Christina <tamar.christina@arm.com>
18508 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
18509 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
18511 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18514 * config/s390/s390.c (s390_set_current_function): Invoke
18515 s390_indirect_branch_settings also if fndecl didn't change.
18517 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
18519 * config/rs6000/rs6000.md (blockage): Set length to zero.
18521 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
18523 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
18525 2018-02-09 Jakub Jelinek <jakub@redhat.com>
18528 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
18529 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
18533 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
18534 PARALLEL incoming that failed vt_get_decl_and_offset check.
18536 PR middle-end/84237
18537 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
18538 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
18540 (get_variable_section): For decls in named .bss* sections pass true as
18541 second argument to bss_initializer_p.
18543 2018-02-09 Marek Polacek <polacek@redhat.com>
18544 Jakub Jelinek <jakub@redhat.com>
18547 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
18548 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
18549 Sync some changes from cxx_fold_indirect_ref.
18551 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
18553 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
18555 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
18556 (BLOCK_INLINE_ENTRY_LABEL): New.
18557 (dwarf2out_var_location): Disregard inline entry markers.
18558 (inline_entry_data): New struct.
18559 (inline_entry_data_hasher): New hashtable type.
18560 (inline_entry_data_hasher::hash): New.
18561 (inline_entry_data_hasher::equal): New.
18562 (inline_entry_data_table): New variable.
18563 (add_high_low_attributes): Add DW_AT_entry_pc and
18564 DW_AT_GNU_entry_view attributes if a pending entry is found
18565 in inline_entry_data_table. Add old entry_pc attribute only
18566 if debug nonbinding markers are disabled.
18567 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
18568 markers are enabled.
18569 (block_within_block_p, dwarf2out_inline_entry): New.
18570 (dwarf2out_finish): Check that no entries remained in
18571 inline_entry_data_table.
18572 * final.c (reemit_insn_block_notes): Handle inline entry notes.
18573 (final_scan_insn, notice_source_line): Likewise.
18574 (rest_of_clean_state): Skip inline entry markers.
18575 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
18577 * gimple.c (gimple_build_debug_inline_entry): New.
18578 * gimple.h (enum gimple_debug_subcode): Add
18579 GIMPLE_DEBUG_INLINE_ENTRY.
18580 (gimple_build_debug_inline_entry): Declare.
18581 (gimple_debug_inline_entry_p): New.
18582 (gimple_debug_nonbind_marker_p): Adjust.
18583 * insn-notes.def (INLINE_ENTRY): New.
18584 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
18585 inline entry marker notes.
18586 (print_insn): Likewise.
18587 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
18588 (INSN_DEBUG_MARKER_KIND): Likewise.
18589 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
18590 * tree-inline.c (expand_call_inline): Build and insert
18591 debug_inline_entry stmt.
18592 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
18593 inline entry blocks early, if nonbind markers are enabled.
18594 (dump_scope_block): Dump fragment info.
18595 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
18596 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
18597 (gimple_build_debug_inline_entry): New.
18598 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
18599 Enable/disable inline entry points too.
18600 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
18601 (DEBUG_INSN): Describe inline entry markers.
18603 * common.opt (gvariable-location-views): New.
18604 (gvariable-location-views=incompat5): New.
18605 * config.in: Rebuilt.
18606 * configure: Rebuilt.
18607 * configure.ac: Test assembler for view support.
18608 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
18609 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
18610 * dwarf2out.c (var_loc_view): New typedef.
18611 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
18612 (dwarf2out_locviews_in_attribute): New.
18613 (dwarf2out_locviews_in_loclist): New.
18614 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
18615 (enum dw_line_info_opcode): Add LI_adv_address.
18616 (struct dw_line_info_table): Add view.
18617 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
18618 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
18619 (zero_view_p): New variable.
18620 (ZERO_VIEW_P): New macro.
18621 (output_asm_line_debug_info): New.
18622 (struct var_loc_node): Add view.
18623 (add_AT_view_list, AT_loc_list): New.
18624 (add_var_loc_to_decl): Add view param. Test it against last.
18625 (new_loc_list): Add view params. Record them.
18626 (AT_loc_list_ptr): Handle loc and view lists.
18627 (view_list_to_loc_list_val_node): New.
18628 (print_dw_val): Handle dw_val_class_view_list.
18629 (size_of_die): Likewise.
18630 (value_format): Likewise.
18631 (loc_list_has_views): New.
18632 (gen_llsym): Set vl_symbol too.
18633 (maybe_gen_llsym, skip_loc_list_entry): New.
18634 (dwarf2out_maybe_output_loclist_view_pair): New.
18635 (output_loc_list): Output view list or entries too.
18636 (output_view_list_offset): New.
18637 (output_die): Handle dw_val_class_view_list.
18638 (output_dwarf_version): New.
18639 (output_compilation_unit_header): Use it.
18640 (output_skeleton_debug_sections): Likewise.
18641 (output_rnglists, output_line_info): Likewise.
18642 (output_pubnames, output_aranges): Update version comments.
18643 (output_one_line_info_table): Output view numbers in asm comments.
18644 (dw_loc_list): Determine current endview, pass it to new_loc_list.
18645 Call maybe_gen_llsym.
18646 (loc_list_from_tree_1): Adjust.
18647 (add_AT_location_description): Create view list attribute if
18648 needed, check it's absent otherwise.
18649 (convert_cfa_to_fb_loc_list): Adjust.
18650 (maybe_emit_file): Call output_asm_line_debug_info for test.
18651 (dwarf2out_var_location): Reset views as needed. Precompute
18652 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
18653 attribute. Set view.
18654 (new_line_info_table): Reset next view.
18655 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
18656 (dwarf2out_source_line): Likewise. Output view resets and labels to
18657 the assembler, or select appropriate line info opcodes.
18658 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
18659 (optimize_string_length): Catch it. Adjust.
18660 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
18661 dw_val_class_view_list, and remove it if no longer needed.
18662 (hash_loc_list): Hash view numbers.
18663 (loc_list_hasher::equal): Compare them.
18664 (optimize_location_lists): Check whether a view list symbol is
18665 needed, and whether the locview attribute is present, and
18666 whether they match. Remove the locview attribute if no longer
18668 (index_location_lists): Call skip_loc_list_entry for test.
18669 (dwarf2out_finish): Call output_asm_line_debug_info for test.
18670 Use output_dwarf_version.
18671 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
18672 (struct dw_val_node): Add val_view_list.
18673 * final.c (SEEN_NEXT_VIEW): New.
18674 (set_next_view_needed): New.
18675 (clear_next_view_needed): New.
18676 (maybe_output_next_view): New.
18677 (final_start_function): Rename to...
18678 (final_start_function_1): ... this. Take pointer to FIRST,
18679 add SEEN parameter. Emit param bindings in the initial view.
18680 (final_start_function): Reintroduce SEEN-less interface.
18681 (final): Rename to...
18682 (final_1): ... this. Take SEEN parameter. Output final pending
18683 next view at the end.
18684 (final): Reintroduce seen-less interface.
18685 (final_scan_insn): Output pending next view before switching
18686 sections or ending a block. Mark the next view as needed when
18687 outputting variable locations. Notify debug backend of section
18688 changes, and of location view changes.
18689 (rest_of_handle_final): Adjust.
18690 * toplev.c (process_options): Autodetect value for debug variable
18691 location views option. Warn on incompat5 without -gdwarf-5.
18692 * doc/invoke.texi (gvariable-location-views): New.
18693 (gvariable-location-views=incompat5): New.
18694 (gno-variable-location-views): New.
18696 2018-02-08 David Malcolm <dmalcolm@redhat.com>
18698 PR tree-optimization/84136
18699 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
18700 that the result of find_edge is non-NULL.
18702 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
18705 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
18706 storing integer register in SImode. Fix cost of 256 and 512
18707 byte aligned SSE register store.
18709 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
18711 * config/i386/i386.c (ix86_multiplication_cost): Fix
18712 multiplication cost for TARGET_AVX512DQ.
18714 2018-02-08 Marek Polacek <polacek@redhat.com>
18716 PR tree-optimization/84238
18717 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
18720 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
18722 PR tree-optimization/84265
18723 * tree-vect-stmts.c (vectorizable_store): Don't treat
18724 VMAT_CONTIGUOUS accesses as grouped.
18725 (vectorizable_load): Likewise.
18727 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
18729 PR tree-optimization/81635
18730 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
18731 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
18732 (test_round_for_mask): New functions.
18733 (wide_int_cc_tests): Call test_round_for_mask.
18734 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
18735 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
18736 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
18737 range returned by get_range_info.
18739 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
18742 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
18743 * symtab.c: Include builtins.h
18744 (symtab_node::output_to_lto_symbol_table_p): Move here
18745 from lto-streamer-out.c:output_symbol_p.
18746 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
18747 (output_symbol_p): Move all logic to symtab.c
18748 (produce_symtab): Update.
18750 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18752 * config/s390/s390-opts.h (enum indirect_branch): Define.
18753 * config/s390/s390-protos.h (s390_return_addr_from_memory)
18754 (s390_indirect_branch_via_thunk)
18755 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
18756 (enum s390_indirect_branch_type): Define.
18757 * config/s390/s390.c (struct s390_frame_layout, struct
18758 machine_function): Remove.
18759 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
18760 (indirect_branch_table_label_no, indirect_branch_table_name):
18762 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
18763 (enum s390_indirect_branch_option): Define.
18764 (s390_return_addr_from_memory): New function.
18765 (s390_handle_string_attribute): New function.
18766 (s390_attribute_table): Add new attribute handler.
18767 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
18768 (s390_indirect_branch_via_thunk): New function.
18769 (s390_indirect_branch_via_inline_thunk): New function.
18770 (s390_function_ok_for_sibcall): When jumping via thunk disallow
18771 sibling call optimization for non z10 compiles.
18772 (s390_emit_call): Force indirect branch target to be a single
18773 register. Add r1 clobber for non-z10 compiles.
18774 (s390_emit_epilogue): Emit return jump via return_use expander.
18775 (s390_reorg): Handle JUMP_INSNs as execute targets.
18776 (s390_option_override_internal): Perform validity checks for the
18777 new command line options.
18778 (s390_indirect_branch_attrvalue): New function.
18779 (s390_indirect_branch_settings): New function.
18780 (s390_set_current_function): Invoke s390_indirect_branch_settings.
18781 (s390_output_indirect_thunk_function): New function.
18782 (s390_code_end): Implement target hook.
18783 (s390_case_values_threshold): Implement target hook.
18784 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
18786 * config/s390/s390.h (struct s390_frame_layout)
18787 (struct machine_function): Move here from s390.c.
18788 (TARGET_INDIRECT_BRANCH_NOBP_RET)
18789 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
18790 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
18791 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
18792 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
18793 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
18794 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
18795 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
18796 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
18797 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
18798 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
18799 (mnemonic attribute): Add values which aren't recognized
18801 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
18802 pattern for branch conversion. Fix mnemonic attribute.
18803 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
18804 indirect branch via thunk if requested.
18805 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
18806 ("*indirect_jump"): Disable for branch conversion using out of
18808 ("indirect_jump_via_thunk<mode>_z10")
18809 ("indirect_jump_via_thunk<mode>")
18810 ("indirect_jump_via_inlinethunk<mode>_z10")
18811 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
18812 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
18813 ("casesi_jump_via_inlinethunk<mode>_z10")
18814 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
18815 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
18816 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
18817 ("*indirect2_jump"): Disable for branch conversion.
18818 ("casesi_jump"): Turn into expander and expand patterns for branch
18820 ("return_use"): New expander.
18821 ("*return"): Emit return via thunk and rename it to ...
18822 ("*return<mode>"): ... this one.
18823 * config/s390/s390.opt: Add new options and and enum for the
18826 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
18828 * lra-constraints.c (match_reload): Unconditionally use
18829 gen_lowpart_SUBREG, rather than selecting between that
18830 and equivalent gen_rtx_SUBREG code.
18832 2018-02-08 Richard Biener <rguenther@suse.de>
18834 PR tree-optimization/84233
18835 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
18836 changed flag instead of boguously re-using phi_inserted.
18838 2018-02-08 Martin Jambor <mjambor@suse.cz>
18840 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
18841 static local variables.
18843 2018-02-08 Richard Biener <rguenther@suse.de>
18845 PR tree-optimization/84278
18846 * tree-vect-stmts.c (vectorizable_store): When looking for
18847 smaller vector types to perform grouped strided loads/stores
18848 make sure the mode is supported by the target.
18849 (vectorizable_load): Likewise.
18851 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
18853 * config/aarch64/aarch64.c (aarch64_components_for_bb):
18854 Increase LDP/STP opportunities by adding adjacent callee-saves.
18856 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
18858 PR rtl-optimization/84068
18859 PR rtl-optimization/83459
18860 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
18862 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
18864 PR tree-optimization/84224
18865 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
18866 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
18867 non-zero arguments.
18869 2018-02-07 Iain Sandoe <iain@codesourcery.com>
18872 * config/rs6000/altivec.md (*restore_world): Remove LR use.
18873 * config/rs6000/predicates.md (restore_world_operation): Adjust op
18874 count, remove one USE.
18876 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
18878 * doc/install.texi (Configuration): Document the
18879 --with-long-double-format={ibm,ieee} PowerPC configuration
18883 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
18884 Convert from define_expand to be define_insn_and_split. Rework
18885 float/double/_Float128 conversions to QI/HI/SImode to work with
18886 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
18887 conversions to QI/HImode types did a store and then a load to
18888 truncate the value. For conversions to VSX registers, don't split
18889 the insn, instead emit the code directly. Use the code iterator
18890 any_fix to combine signed and unsigned conversions.
18891 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
18892 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
18893 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
18894 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
18895 (fix_<mode>di2_hw): Likewise.
18896 (fixuns_<mode>di2_hw): Likewise.
18897 (fix_<mode>si2_hw): Likewise.
18898 (fixuns_<mode>si2_hw): Likewise.
18899 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
18900 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
18901 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
18902 fix<uns>_trunc<SFDF:mode>si2_p8.
18903 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
18904 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
18905 (fix<uns>_<mode>_mem): Likewise.
18906 (fctiw<u>z_<mode>_mem): Likewise.
18907 (fix<uns>_<mode>_mem): Likewise.
18908 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
18909 the register allocator from doing a direct move to the GPRs to do
18910 a store, and instead use the ISA 3.0 store byte/half-word from
18911 vector register instruction. For IEEE 128-bit floating point,
18912 also optimize stores of 32-bit ints.
18913 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
18915 2018-02-07 Alan Hayward <alan.hayward@arm.com>
18917 * genextract.c (push_pathstr_operand): New function to support
18919 (walk_rtx): Call push_pathstr_operand.
18920 (print_path): Support [a-zA-Z].
18922 2018-02-07 Richard Biener <rguenther@suse.de>
18924 PR tree-optimization/84037
18925 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
18926 (cse_and_gimplify_to_preheader): Declare.
18927 (vect_get_place_in_interleaving_chain): Likewise.
18928 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
18930 (_loop_vec_info::~_loop_vec_info): Delete it.
18931 (cse_and_gimplify_to_preheader): New function.
18932 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
18933 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
18934 (vectorizable_load): Likewise. For grouped stores always base
18935 the IV on the first element.
18936 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
18937 condition before gimplifying.
18939 2018-02-07 Jakub Jelinek <jakub@redhat.com>
18941 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
18942 *DIV_EXPR and *MOD_EXPR.
18944 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
18947 * config/i386/i386.c (ix86_option_override_internal): Mask out
18948 the CF_SET bit when checking -fcf-protection.
18950 2018-02-07 Tom de Vries <tom@codesourcery.com>
18953 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
18956 2018-02-07 Richard Biener <rguenther@suse.de>
18958 PR tree-optimization/84204
18959 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
18962 PR tree-optimization/84205
18963 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
18964 special-case isl_ast_op_zdiv_r.
18966 PR tree-optimization/84223
18967 * graphite-scop-detection.c (gather_bbs::before_dom_children):
18968 Only add conditions from within the region.
18969 (gather_bbs::after_dom_children): Adjust.
18971 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
18974 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
18975 * config/avr/avr.md: Only post-reload split REG-REG moves if
18976 either register is GENERAL_REG_P.
18978 2018-02-07 Jakub Jelinek <jakub@redhat.com>
18980 PR tree-optimization/84235
18981 * tree-ssa-scopedtables.c
18982 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
18983 if the subtraction is performed in floating point type where NaNs are
18984 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
18985 build 1. Formatting fix.
18987 2018-02-06 Jakub Jelinek <jakub@redhat.com>
18990 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
18991 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
18992 and skip it regardless of bb boundaries. Use CALL_P macro,
18993 don't test INSN_P (insn) together with CALL_P or JUMP_P check
18994 unnecessarily, formatting fix.
18996 2018-02-06 Michael Collison <michael.collison@arm.com>
18998 * config/arm/thumb2.md:
18999 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
19000 (*thumb_mov_notscc): Ditto.
19002 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
19005 * config/rs6000/rs6000.md (su code attribute): Use "u" for
19006 unsigned_fix, not "s".
19008 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19010 * configure.ac (gcc_fn_eh_frame_ro): New function.
19011 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
19012 correct .eh_frame permissions.
19013 * configure: Regenerate.
19015 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
19017 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
19018 irrelevant options.
19020 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19022 * config/rs6000/rs6000.c (rs6000_option_override_internal):
19023 Display warning message for -mno-speculate-indirect-jumps.
19025 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
19027 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
19029 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
19031 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
19033 PR tree-optimization/84225
19034 * tree-eh.c (find_trapping_overflow): Only call
19035 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
19037 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
19040 * config/i386/i386.c: Reimplement the check of possible options
19041 -mibt/-mshstk conbination. Change error messages.
19042 * doc/invoke.texi: Fix a typo: remove extra '='.
19044 2018-02-06 Marek Polacek <polacek@redhat.com>
19046 PR tree-optimization/84228
19047 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
19049 2018-02-06 Tamar Christina <tamar.christina@arm.com>
19052 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
19053 emitted arch directives.
19054 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
19055 __ARM_FEATURE_COPROC before changing architectures.
19057 2018-02-06 Richard Biener <rguenther@suse.de>
19059 * config/i386/i386.c (print_reg): Fix typo.
19060 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
19062 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
19064 * configure: Regenerate.
19066 2018-02-05 Martin Sebor <msebor@redhat.com>
19068 PR tree-optimization/83369
19069 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
19072 2018-02-05 Martin Liska <mliska@suse.cz>
19074 * doc/invoke.texi: Cherry-pick upstream r323995.
19076 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
19078 * ira.c (ira_init_register_move_cost): Adjust comment.
19080 2018-02-05 Martin Liska <mliska@suse.cz>
19082 PR gcov-profile/84137
19083 * doc/gcov.texi: Fix typo in documentation.
19085 2018-02-05 Martin Liska <mliska@suse.cz>
19087 PR gcov-profile/83879
19088 * doc/gcov.texi: Document necessity of --dynamic-list-data when
19089 using dlopen functionality.
19091 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
19093 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
19094 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
19095 _mm_maskz_range_ss, _mm_mask_range_round_ss,
19096 _mm_maskz_range_round_ss): New intrinsics.
19097 (__builtin_ia32_rangesd128_round)
19098 (__builtin_ia32_rangess128_round): Remove.
19099 (__builtin_ia32_rangesd128_mask_round,
19100 __builtin_ia32_rangess128_mask_round): New builtins.
19101 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
19102 __builtin_ia32_rangess128_round): Remove.
19103 (__builtin_ia32_rangesd128_mask_round,
19104 __builtin_ia32_rangess128_mask_round): New builtins.
19105 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
19106 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
19107 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
19108 "<round_saeonly_constraint>")): Changed to ...
19109 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
19110 "<round_saeonly_scalar_constraint>")): ... this.
19111 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
19112 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
19113 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
19114 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
19115 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
19117 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
19119 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
19121 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
19122 Remove all values except native, 8540 and 8548.
19124 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
19126 * config/i386/i386.c (ix86_output_function_return): Pass
19127 INVALID_REGNUM, instead of -1, as invalid register number to
19128 indirect_thunk_name and output_indirect_thunk.
19130 2018-02-02 Julia Koval <julia.koval@intel.com>
19132 * config.gcc: Add -march=icelake.
19133 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
19134 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
19135 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
19136 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
19137 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
19138 (processor_target_table): Add icelake.
19139 (ix86_option_override_internal): Handle new PTAs.
19140 (get_builtin_code_for_version): Handle icelake.
19141 (M_INTEL_COREI7_ICELAKE): New.
19142 (fold_builtin_cpu): Handle icelake.
19143 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
19144 * doc/invoke.texi: Add -march=icelake.
19146 2018-02-02 Julia Koval <julia.koval@intel.com>
19148 * config/i386/i386.c (ix86_option_override_internal): Change flags type
19149 to wide_int_bitmask.
19150 * wide-int-bitmask.h: New.
19152 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
19155 * config/i386/i386.md: Replace Pmode with word_mode in
19156 builtin_setjmp_setup and builtin_longjmp to support x32.
19158 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
19162 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
19164 (rs6000_supported_cpu_names): New static variable.
19165 (linux_cpu_translation_table): Likewise.
19166 (elf_platform) <cpu>: Define new static variable and use it.
19167 Translate kernel AT_PLATFORM name to canonical name if needed.
19168 Error if platform name is unknown.
19170 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
19173 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
19175 2018-02-01 Jeff Law <law@redhat.com>
19178 * config/i386/i386.c (release_scratch_register_on_entry): Add new
19179 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
19180 the scratch if RELEASE_VIA_POP is false.
19181 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
19182 If we have to save a temporary register, decrement SIZE appropriately.
19183 Pass new arguments to release_scratch_register_on_entry.
19184 (ix86_adjust_stack_and_probe): Likewise.
19185 (ix86_emit_probe_stack_range): Pass new arguments to
19186 release_scratch_register_on_entry.
19188 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
19190 PR rtl-optimization/84157
19191 * combine.c (change_zero_ext): Use REG_P predicate in
19192 front of HARD_REGISTER_P predicate.
19194 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
19196 * config/avr/avr.c (avr_option_override): Move disabling of
19197 -fdelete-null-pointer-checks to...
19198 * common/config/avr/avr-common.c (avr_option_optimization_table):
19201 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19203 PR tree-optimization/81635
19204 * tree-data-ref.c (split_constant_offset_1): For types that
19205 wrap on overflow, try to use range info to prove that wrapping
19208 2018-02-01 Renlin Li <renlin.li@arm.com>
19211 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
19212 TAILCALL_ADDR_REGS.
19213 (aarch64_register_move_cost): Likewise.
19214 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
19215 TAILCALL_ADDR_REGS.
19216 (REG_CLASS_NAMES): Likewise.
19217 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
19218 TAILCALL_ADDR_REGS. Remove IP registers.
19219 * config/aarch64/aarch64.md (Ucs): Update register constraint.
19221 2018-02-01 Richard Biener <rguenther@suse.de>
19223 * domwalk.h (dom_walker::dom_walker): Add additional constructor
19224 for specifying RPO order and allow NULL for that.
19225 * domwalk.c (dom_walker::dom_walker): Likewise.
19226 (dom_walker::walk): Handle NULL RPO order.
19227 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
19229 (rewrite_update_dom_walker): Likewise.
19230 (mark_def_dom_walker): Likewise.
19232 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19234 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
19235 (aarch64_maybe_expand_sve_subreg_move): Declare.
19236 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
19237 * config/aarch64/predicates.md (aarch64_any_register_operand): New
19239 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
19240 that are semantically a reverse operation.
19241 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
19242 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
19243 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
19245 (aarch64_can_change_mode_class): For big-endian, forbid changes
19246 between two SVE modes if they have different element sizes.
19248 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19250 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
19251 the TImode handling for big-endian targets.
19253 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19255 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
19256 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
19258 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
19259 Remove BSWAP handing for big-endian targets and use the form of
19260 LD1RQ appropariate for the mode.
19262 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19264 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
19265 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
19266 duplicated element.
19268 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19271 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
19272 check for operands that need to go through aarch64_sve_reload_be.
19274 2018-02-01 Jakub Jelinek <jakub@redhat.com>
19276 PR tree-optimization/81661
19277 PR tree-optimization/84117
19278 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
19279 * tree-eh.c: Include gimplify.h.
19280 (find_trapping_overflow, replace_trapping_overflow,
19281 rewrite_to_non_trapping_overflow): New functions.
19282 * tree-vect-loop.c: Include tree-eh.h.
19283 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
19284 * tree-data-ref.c: Include tree-eh.h.
19285 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
19287 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
19289 PR rtl-optimization/84123
19290 * combine.c (change_zero_ext): Check if hard register satisfies
19291 can_change_dest_mode before calling gen_lowpart_SUBREG.
19293 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
19296 * ira.c (ira_init_register_move_cost): Remove assert.
19298 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19300 PR rtl-optimization/84071
19301 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
19302 * doc/tm.texi: Regenerate.
19304 2018-01-31 Richard Biener <rguenther@suse.de>
19306 PR tree-optimization/84132
19307 * tree-data-ref.c (analyze_miv_subscript): Properly
19308 check whether evolution_function_is_affine_multivariate_p
19309 before calling gcd_of_steps_may_divide_p.
19311 2018-01-31 Julia Koval <julia.koval@intel.com>
19314 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
19315 * config/i386/i386.md (rdpid_rex64) New.
19316 (rdpid): Make 32bit only.
19318 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
19321 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
19322 an IDENTIFIER_NODE for FUNCTION_TYPE's.
19324 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19327 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
19329 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
19331 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19333 PR rtl-optimization/84071
19334 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
19335 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
19337 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
19339 * config/arc/arc.c (arc_handle_aux_attribute): New function.
19340 (arc_attribute_table): Add 'aux' attribute.
19341 (arc_in_small_data_p): Consider aux like variables.
19342 (arc_is_aux_reg_p): New function.
19343 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
19344 (arc_get_aux_arg): New function.
19345 (prepare_move_operands): Handle aux-register access.
19346 (arc_handle_aux_attribute): New function.
19347 * doc/extend.texi (ARC Variable attributes): Add subsection.
19349 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
19351 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
19352 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
19353 (arc_attribute_table): Add 'uncached' attribute.
19354 (arc_print_operand): Print '.di' flag for uncached memory
19356 (arc_in_small_data_p): Do not consider for small data the uncached
19358 (arc_is_uncached_mem_p): New function.
19359 * config/arc/predicates.md (compact_store_memory_operand): Check
19360 for uncached memory accesses.
19361 (nonvol_nonimm_operand): Likewise.
19362 * doc/extend.texi (ARC Type Attribute): New subsection.
19364 2018-01-31 Jakub Jelinek <jakub@redhat.com>
19367 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
19368 falign-loops=): Add Optimization flag.
19370 2018-01-30 Jeff Law <law@redhat.com>
19373 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
19374 INT_REGISTERS_SAVED. Check it prior to calling
19375 get_scratch_register_on_entry.
19376 (ix86_adjust_stack_and_probe): Similarly.
19377 (ix86_emit_probe_stack_range): Similarly.
19378 (ix86_expand_prologue): Corresponding changes.
19380 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19383 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
19384 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
19386 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
19389 * lra-constraints.c (curr_insn_transform): Process AND in the
19392 2018-01-30 Jakub Jelinek <jakub@redhat.com>
19394 PR rtl-optimization/83986
19395 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
19396 dependence against last_pending_memory_flush in addition to
19397 pending_jump_insns.
19399 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
19401 PR tree-optimization/81611
19402 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
19405 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19408 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
19411 2018-01-30 Richard Biener <rguenther@suse.de>
19412 Jakub Jelinek <jakub@redhat.com>
19414 PR tree-optimization/84111
19415 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
19416 inner loops added during recursion, as they don't have up-to-date
19419 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
19422 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
19423 (can_inline_edge_by_limits_p): ... here.
19424 (can_early_inline_edge_p, check_callers,
19425 update_caller_keys, update_callee_keys, recursive_inlining,
19426 add_new_edges_to_heap, speculation_useful_p,
19427 inline_small_functions,
19428 inline_small_functions, flatten_function,
19429 inline_to_all_callers_1): Update.
19431 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
19433 * profile-count.c (profile_count::combine_with_ipa_count): Handle
19436 2018-01-30 Richard Biener <rguenther@suse.de>
19438 PR tree-optimization/83008
19439 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
19440 invariant and constant vector uses in stmts when they need
19441 more than one stmt.
19443 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19446 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
19447 * configure: Regenerate.
19449 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
19451 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
19453 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
19454 Use gen_rtx_REG rather than gen_lowpart.
19456 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
19458 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
19459 rather than 0 when creating partial subregs.
19461 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
19463 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
19466 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
19469 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
19470 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
19471 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
19472 flags. This restores the settings used before the 2017-07-24.
19473 Turning off pre increment/decrement/modify allows IVOPTS to
19474 optimize DF/SF loops where the index is an int.
19476 2018-01-29 Richard Biener <rguenther@suse.de>
19477 Kelvin Nilsen <kelvin@gcc.gnu.org>
19480 * tree-vect-stmts.c (vectorizable_call): Don't call
19481 targetm.vectorize_builtin_md_vectorized_function if callee is
19484 2018-01-22 Carl Love <cel@us.ibm.com>
19486 * doc/extend.tex: Fix typo in second arg in
19487 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
19489 2018-01-29 Richard Biener <rguenther@suse.de>
19491 PR tree-optimization/84086
19492 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
19493 (flush_ssaname_freelist): When SSA names were released reset
19494 the SCEV hash table.
19496 2018-01-29 Richard Biener <rguenther@suse.de>
19498 PR tree-optimization/84057
19499 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
19500 removed paths when removing edges.
19502 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
19504 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
19505 -mfunction-return=@var{choice}.
19507 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
19509 PR diagnostic/84034
19510 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
19511 Handle CR like TAB.
19512 (layout::print_source_line): Likewise.
19513 (test_get_line_width_without_trailing_whitespace): Add test cases.
19515 2018-01-27 Jakub Jelinek <jakub@redhat.com>
19517 PR middle-end/84040
19518 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
19521 2018-01-26 Jim Wilson <jimw@sifive.com>
19523 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
19525 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
19528 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19530 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
19531 and CMP + SUB-immediate -> SUBS.
19533 2018-01-26 Martin Sebor <msebor@redhat.com>
19535 PR tree-optimization/83896
19536 * tree-ssa-strlen.c (get_string_len): Rename...
19537 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
19538 Avoid assuming length is constant.
19539 (handle_char_store): Use HOST_WIDE_INT for string length.
19541 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
19544 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
19545 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
19547 2018-01-26 Richard Biener <rguenther@suse.de>
19549 PR rtl-optimization/84003
19550 * dse.c (record_store): Only record redundant stores when
19551 the earlier store aliases at least all accesses the later one does.
19553 2018-01-26 Jakub Jelinek <jakub@redhat.com>
19555 PR rtl-optimization/83985
19556 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
19557 REG_CFA_RESTORE insns.
19558 (delete_unmarked_insns): Don't ignore separate shrink wrapping
19559 REG_CFA_RESTORE insns here.
19562 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
19563 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
19565 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19567 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
19568 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
19569 (arc_init): Likewise.
19570 (arc_override_options): Likewise.
19571 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
19573 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
19575 * config/arc/arc.h (TARGET_DBNZ): Define.
19576 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
19577 properly set the tune attribute.
19578 (dbnz): Use TARGET_DBNZ guard.
19579 * config/arc/arc.opt (mtune): Add core3 option.
19581 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19583 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
19584 recognize new pic like addresses.
19585 (arc_delegitimize_address): Clean up.
19587 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19589 * config/arc/arc-arches.def: Option mrf16 valid for all
19591 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
19592 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
19593 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
19594 * config/arc/arc-tables.opt: Regenerate.
19595 * config/arc/arc.c (arc_conditional_register_usage): Handle
19596 reduced register file case.
19597 (arc_file_start): Set must have build attributes.
19598 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
19599 mrf16 option value.
19600 * config/arc/arc.opt (mrf16): Add new option.
19601 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
19602 * config/arc/genmultilib.awk: Handle new mrf16 option.
19603 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
19604 * config/arc/t-multilib: Regenerate.
19605 * doc/invoke.texi (ARC Options): Document mrf16 option.
19607 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19609 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
19610 * config/arc/arc.c (arc_handle_secure_attribute): New function.
19611 (arc_attribute_table): Add 'secure_call' attribute.
19612 (arc_print_operand): Print secure call operand.
19613 (arc_function_ok_for_sibcall): Don't optimize tail calls when
19615 (arc_is_secure_call_p): New function. * config/arc/arc.md
19616 (call_i): Add support for sjli instruction.
19617 (call_value_i): Likewise.
19618 * config/arc/constraints.md (Csc): New constraint.
19620 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19621 John Eric Martin <John.Martin@emmicro-us.com>
19623 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
19624 * config/arc/arc.c (_arc_jli_section): New struct.
19625 (arc_jli_section): New type.
19626 (rc_jli_sections): New static variable.
19627 (arc_handle_jli_attribute): New function.
19628 (arc_attribute_table): Add jli_always and jli_fixed attribute.
19629 (arc_file_end): New function.
19630 (TARGET_ASM_FILE_END): Define.
19631 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
19632 (arc_add_jli_section): New function.
19633 (jli_call_scan): Likewise.
19634 (arc_reorg): Call jli_call_scan.
19635 (arc_output_addsi): Remove 'S' from printing asm operand.
19636 (arc_is_jli_call_p): New function.
19637 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
19639 (movhi_insn): Likewise.
19640 (movsi_insn): Likewise.
19641 (movsi_set_cc_insn): Likewise.
19642 (loadqi_update): Likewise.
19643 (load_zeroextendqisi_update): Likewise.
19644 (load_signextendqisi_update): Likewise.
19645 (loadhi_update): Likewise.
19646 (load_zeroextendhisi_update): Likewise.
19647 (load_signextendhisi_update): Likewise.
19648 (loadsi_update): Likewise.
19649 (loadsf_update): Likewise.
19650 (movsicc_insn): Likewise.
19651 (bset_insn): Likewise.
19652 (bxor_insn): Likewise.
19653 (bclr_insn): Likewise.
19654 (bmsk_insn): Likewise.
19655 (bicsi3_insn): Likewise.
19656 (cmpsi_cc_c_insn): Likewise.
19657 (movsi_ne): Likewise.
19658 (movsi_cond_exec): Likewise.
19659 (clrsbsi2): Likewise.
19660 (norm_f): Likewise.
19670 (call_i): Remove 'S' asm letter, add jli instruction.
19671 (call_value_i): Likewise.
19672 * config/arc/arc.op (mjli-always): New option.
19673 * config/arc/constraints.md (Cji): New constraint.
19674 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
19676 (subsf3_fpx): Likewise.
19677 (mulsf3_fpx): Likewise.
19678 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
19680 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
19681 function attrbutes.
19682 * doc/invoke.texi (ARC): Document mjli-always option.
19684 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
19686 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
19687 avoid addition with 0 and use incw and decw where possible.
19689 2018-01-26 Richard Biener <rguenther@suse.de>
19691 PR tree-optimization/81082
19692 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
19693 association if it requires casting to unsigned.
19694 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
19695 from fold_plusminus_mult_expr to catch important cases late when
19696 range info is available.
19698 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19700 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
19701 * configure.ac (hidden_linkonce): New test.
19702 * configure: Regenerate.
19703 * config.in: Regenerate.
19705 2018-01-26 Julia Koval <julia.koval@intel.com>
19707 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
19708 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
19709 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
19710 _mm_mask_bitshuffle_epi64_mask): Fix type.
19711 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
19712 USI_FTYPE_V4DI_V4DI_USI): Remove.
19713 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
19714 __builtin_ia32_vpshufbitqmb256_mask,
19715 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
19716 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
19717 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
19719 2018-01-26 Alan Modra <amodra@gmail.com>
19722 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
19723 UNSPEC_VBPERMQ. Sort other unspecs.
19725 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
19727 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
19729 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
19731 PR middle-end/83055
19732 * predict.c (drop_profile): Do not push/pop cfun; update also
19734 (handle_missing_profiles): Fix logic looking for zero profiles.
19736 2018-01-25 Jakub Jelinek <jakub@redhat.com>
19738 PR middle-end/83977
19739 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
19740 on functions with #pragma omp declare simd or functions with simd
19742 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
19743 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
19744 Remove trailing \n from warning_at calls.
19746 2018-01-25 Tom de Vries <tom@codesourcery.com>
19749 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
19750 for neutered workers.
19752 2018-01-24 Joseph Myers <joseph@codesourcery.com>
19755 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
19756 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
19758 2018-01-24 Jeff Law <law@redhat.com>
19761 * i386.c (get_probe_interval): Move to earlier point.
19762 (ix86_compute_frame_layout): If -fstack-clash-protection and
19763 the frame is larger than the probe interval, then use pushes
19764 to save registers rather than reg->mem moves.
19765 (ix86_expand_prologue): Remove conditional for int_registers_saved
19768 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
19771 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
19772 min/max for never referenced object.
19774 2018-01-24 Jakub Jelinek <jakub@redhat.com>
19776 PR middle-end/83977
19777 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
19779 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
19780 attributes from DECL_ATTRIBUTES (decl) without affecting
19781 DECL_ATTRIBUTES (current_function_decl).
19782 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
19783 functions with non-NULL DECL_ABSTRACT_ORIGIN.
19785 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
19787 PR tree-optimization/83979
19788 * fold-const.c (fold_comparison): Use constant_boolean_node
19789 instead of boolean_{true,false}_node.
19791 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
19793 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
19796 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19798 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
19799 Simplify the clause that sets the length attribute.
19800 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
19801 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
19802 clause that sets the length attribute.
19803 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
19805 2018-01-24 Tom de Vries <tom@codesourcery.com>
19808 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
19809 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
19810 Add strict parameter.
19811 (prevent_branch_around_nothing): Insert dummy insn between branch to
19812 label and label with no ptx insn inbetween.
19813 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
19815 2018-01-24 Tom de Vries <tom@codesourcery.com>
19818 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
19819 for neutered threads in warp.
19820 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
19822 2018-01-24 Richard Biener <rguenther@suse.de>
19824 PR tree-optimization/83176
19825 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
19828 2018-01-24 Richard Biener <rguenther@suse.de>
19830 PR tree-optimization/82819
19831 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
19832 code generating pluses that are no-ops in the target precision.
19834 2018-01-24 Richard Biener <rguenther@suse.de>
19836 PR middle-end/84000
19837 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
19839 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
19841 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
19842 to merge probabilities.
19843 * predict.c (probably_never_executed): Also mark as cold functions
19844 with global 0 profile and guessed local profile.
19845 * profile-count.c (profile_probability::combine_with_count): New
19847 * profile-count.h (profile_probability::operator*,
19848 profile_probability::operator*=, profile_probability::operator/,
19849 profile_probability::operator/=): Reduce precision to adjusted
19850 and set value to guessed on contradictory divisions.
19851 (profile_probability::combine_with_freq): Remove.
19852 (profile_probability::combine_wiht_count): Declare.
19853 (profile_count::force_nonzero):: Set to adjusted.
19854 (profile_count::probability_in):: Set quality to adjusted.
19855 * tree-ssa-tail-merge.c (replace_block_by): Use
19856 combine_with_count.
19858 2018-01-23 Andrew Waterman <andrew@sifive.com>
19859 Jim Wilson <jimw@sifive.com>
19861 * config/riscv/riscv.c (riscv_stack_boundary): New.
19862 (riscv_option_override): Set riscv_stack_boundary. Handle
19863 riscv_preferred_stack_boundary_arg.
19864 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
19865 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
19866 (STACK_BOUNDARY): Set to riscv_stack_boundary.
19867 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
19868 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
19869 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
19871 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
19874 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
19875 of struct ix86_frame.
19876 (ix86_expand_epilogue): Likewise. Add a local variable for
19877 the reg_save_offset field in struct ix86_frame.
19879 2018-01-23 Bin Cheng <bin.cheng@arm.com>
19881 PR tree-optimization/82604
19882 * tree-loop-distribution.c (enum partition_kind): New enum item
19883 PKIND_PARTIAL_MEMSET.
19884 (partition_builtin_p): Support above new enum item.
19885 (generate_code_for_partition): Ditto.
19886 (compute_access_range): Differentiate cases that equality can be
19887 proven at all loops, the innermost loops or no loops.
19888 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
19889 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
19890 (finalize_partitions, distribute_loop): Don't fuse partition of
19891 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
19892 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
19893 parloop is enabled.
19895 2018-01-23 Martin Liska <mliska@suse.cz>
19897 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
19898 order to ignore the predictor.
19899 (PRED_POLYMORPHIC_CALL): Likewise.
19900 (PRED_RECURSIVE_CALL): Likewise.
19902 2018-01-23 Martin Liska <mliska@suse.cz>
19904 * tree-profile.c (tree_profiling): Print function header to
19905 aware reader which function we are working on.
19906 * value-prof.c (gimple_find_values_to_profile): Do not print
19907 not interesting value histograms.
19909 2018-01-23 Martin Liska <mliska@suse.cz>
19911 * profile-count.h (enum profile_quality): Add
19912 profile_uninitialized as the first value. Do not number values
19913 as they are zero based.
19914 (profile_count::verify): Update sanity check.
19915 (profile_probability::verify): Likewise.
19917 2018-01-23 Nathan Sidwell <nathan@acm.org>
19919 * doc/invoke.texi (ffor-scope): Deprecate.
19921 2018-01-23 David Malcolm <dmalcolm@redhat.com>
19923 PR tree-optimization/83510
19924 * domwalk.c (set_all_edges_as_executable): New function.
19925 (dom_walker::dom_walker): Convert bool param
19926 "skip_unreachable_blocks" to enum reachability. Move setup of
19927 edge flags to set_all_edges_as_executable and only do it when
19928 reachability is REACHABLE_BLOCKS.
19929 * domwalk.h (enum dom_walker::reachability): New enum.
19930 (dom_walker::dom_walker): Convert bool param
19931 "skip_unreachable_blocks" to enum reachability.
19932 (set_all_edges_as_executable): New decl.
19933 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
19934 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
19936 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
19937 but converting true to REACHABLE_BLOCKS.
19938 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
19940 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
19941 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
19942 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
19944 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
19945 if check_all_array_refs will be called.
19947 2018-01-23 David Malcolm <dmalcolm@redhat.com>
19949 * tree.c (selftest::test_location_wrappers): Add more test
19952 2018-01-23 David Malcolm <dmalcolm@redhat.com>
19954 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
19955 (selftest::test_bit_in_range): Likewise.
19957 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
19960 * doc/sourcebuild.texi (vect_float): Say that the selector
19961 only describes the situation when -funsafe-math-optimizations is on.
19962 (vect_float_strict): Document.
19964 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
19966 PR tree-optimization/83965
19967 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
19968 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
19969 instead of checking only for a reduction.
19970 (vect_recog_widen_sum_pattern): Likewise.
19972 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
19974 * predict.c (probably_never_executed): Only use precise profile info.
19975 (compute_function_frequency): Skip after inlining hack since we now
19976 have quality checking.
19978 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
19980 * profile-count.h (profile_probability::very_unlikely,
19981 profile_probability::unlikely, profile_probability::even): Set
19982 precision to guessed.
19984 2018-01-23 Richard Biener <rguenther@suse.de>
19986 PR tree-optimization/83963
19987 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
19988 Properly terminate dominator walk when crossing the exit edge not
19989 when visiting its source block.
19991 2018-01-23 Jakub Jelinek <jakub@redhat.com>
19994 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
19995 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
19997 2018-01-22 Jakub Jelinek <jakub@redhat.com>
19999 PR tree-optimization/83957
20000 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
20001 semicolon after for body surrounded by braces.
20003 PR tree-optimization/83081
20004 * profile-count.h (profile_probability::split): New method.
20005 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
20006 Use profile_probability::split.
20007 (do_compare_rtx_and_jump): Fix adjustment of probabilities
20008 when splitting a single conditional jump into 2.
20010 2018-01-22 David Malcolm <dmalcolm@redhat.com>
20012 PR tree-optimization/69452
20013 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
20016 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20018 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
20019 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
20020 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
20022 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20024 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
20026 * config/rl78/rl78.md (movdi): New define_expand.
20027 * config/rl78/rl78.c (rl78_split_movdi): New function.
20029 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
20032 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
20034 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
20035 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
20036 128-bit to produce an UNSPEC move to get the double word with the
20037 signbit and then a shift directly to do signbit.
20038 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
20039 implementation with a new version that just does either a direct
20040 move or a regular move. Move memory interface to separate insns.
20041 Move insns so they are next to the expander.
20042 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
20043 with signbit move. Split big and little endian case.
20044 (signbit<mode>2_dm_mem_le): Likewise.
20045 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
20046 (signbit<mode>2_dm2): Likewise.
20048 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20050 * config/rl78/rl78.md (anddi3): New define_expand.
20052 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20054 * config/rl78/rl78.md (umindi3): New define_expand.
20056 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20058 * config/rl78/rl78.md (smindi3): New define_expand.
20060 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20062 * config/rl78/rl78.md (smaxdi3): New define_expand.
20064 2018-01-22 Carl Love <cel@us.ibm.com>
20066 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
20067 LVX_V1TI): Add macro expansion.
20068 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
20069 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
20070 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
20071 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
20072 Change check to determine if the instruction is a byte reversing
20073 entry. Fix typo in comment.
20074 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
20075 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
20076 Add def_builtin calls for new builtins.
20077 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
20078 Add define_insn expansion.
20080 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20082 * config/rl78/rl78.md (umaxdi3): New define_expand.
20084 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20086 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
20087 for non-QImode registers.
20089 2018-01-22 Richard Biener <rguenther@suse.de>
20091 PR tree-optimization/83963
20092 * graphite-scop-detection.c (scop_detection::get_sese): Delay
20093 including the loop exit block.
20094 (scop_detection::merge_sese): Likewise.
20095 (scop_detection::add_scop): Do it here instead.
20097 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20099 * doc/sourcebuild.texi (arm_softfloat): Document.
20101 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
20104 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
20105 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
20106 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
20108 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20109 David Edelsohn <dje.gcc@gmail.com>
20112 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20113 Change "crset eq" to "crset 2".
20114 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20115 (*call_indirect_aix<mode>_nospec): Likewise.
20116 (*call_value_indirect_aix<mode>_nospec): Likewise.
20117 (*call_indirect_elfv2<mode>_nospec): Likewise.
20118 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
20119 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
20120 change assembly output from . to $.
20121 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20122 (indirect_jump<mode>_nospec): Change assembly output from . to $.
20123 (*tablejump<mode>_internal1_nospec): Likewise.
20125 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
20128 * config/sh/sh_optimize_sett_clrt.cc:
20129 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
20131 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
20133 PR tree-optimization/83940
20134 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
20135 offset_dt to vect_constant_def rather than vect_unknown_def_type.
20136 (vect_check_load_store_mask): Add a mask_dt_out parameter and
20137 use it to pass back the definition type.
20138 (vect_check_store_rhs): Likewise rhs_dt_out.
20139 (vect_build_gather_load_calls): Add a mask_dt argument and use
20140 it instead of a call to vect_is_simple_use.
20141 (vectorizable_store): Update calls to vect_check_load_store_mask
20142 and vect_check_store_rhs. Use the dt returned by the latter instead
20143 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
20144 instead of calls to vect_is_simple_use. Pass the scalar rather
20145 than the vector operand to vect_is_simple_use when handling
20146 second and subsequent copies of an rhs value.
20147 (vectorizable_load): Update calls to vect_check_load_store_mask
20148 and vect_build_gather_load_calls. Use the cached mask_dt and
20149 gs_info.offset_dt instead of calls to vect_is_simple_use.
20151 2018-01-20 Jakub Jelinek <jakub@redhat.com>
20153 PR middle-end/83945
20154 * tree-emutls.c: Include gimplify.h.
20155 (lower_emutls_2): New function.
20156 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
20157 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
20158 it before further processing.
20161 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
20162 UINTVAL (trueop1) instead of INTVAL (op1).
20164 2018-01-19 Jakub Jelinek <jakub@redhat.com>
20168 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
20169 INCOMING_FRAME_SP_OFFSET if not defined.
20170 (scan_trace): Add ENTRY argument. If true and
20171 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
20172 emit a note to adjust the CFA offset.
20173 (create_cfi_notes): Adjust scan_trace callers.
20174 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
20175 INCOMING_FRAME_SP_OFFSET in the CIE.
20176 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
20177 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
20179 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
20180 * doc/tm.texi: Regenerated.
20182 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
20184 PR rtl-optimization/83147
20185 * lra-constraints.c (remove_inheritance_pseudos): Use
20186 lra_substitute_pseudo_within_insn.
20188 2018-01-19 Tom de Vries <tom@codesourcery.com>
20189 Cesar Philippidis <cesar@codesourcery.com>
20192 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
20194 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
20197 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
20198 spaces for function labels.
20200 2018-01-19 Martin Liska <mliska@suse.cz>
20202 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
20203 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
20204 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
20205 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
20206 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
20207 (PRED_CONST_RETURN): Change from 69 to 65.
20208 (PRED_NULL_RETURN): Change from 91 to 71.
20209 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
20210 (PRED_LOOP_GUARD): Change from 66 to 73.
20212 2018-01-19 Martin Liska <mliska@suse.cz>
20214 * predict.c (predict_insn_def): Add new assert.
20215 (struct branch_predictor): Change type to signed integer.
20216 (test_prediction_value_range): Amend test to cover
20217 PROB_UNINITIALIZED.
20218 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
20219 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
20220 (PRED_LOOP_ITERATIONS_MAX): Likewise.
20221 (PRED_LOOP_IV_COMPARE): Likewise.
20222 * predict.h (PROB_UNINITIALIZED): Define new constant.
20224 2018-01-19 Martin Liska <mliska@suse.cz>
20226 * predict.c (dump_prediction): Add new format for
20227 analyze_brprob.py script which is enabled with -details
20229 * profile-count.h (precise_p): New function.
20231 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
20233 PR tree-optimization/83922
20234 * tree-vect-loop.c (vect_verify_full_masking): Return false if
20235 there are no statements that need masking.
20236 (vect_active_double_reduction_p): New function.
20237 (vect_analyze_loop_operations): Use it when handling phis that
20238 are not in the loop header.
20240 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
20242 PR tree-optimization/83914
20243 * tree-vect-loop.c (vectorizable_induction): Don't convert
20244 init_expr or apply the peeling adjustment for inductions
20245 that are nested within the vectorized loop.
20247 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20249 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
20252 2018-01-18 Jakub Jelinek <jakub@redhat.com>
20256 * function.h (gimplify_parameters): Add gimple_seq * argument.
20257 * function.c: Include gimple.h and options.h.
20258 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
20259 for the added local temporaries if needed.
20260 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
20261 if there are any parameter cleanups, wrap whole body into a
20262 try/finally with the cleanups.
20264 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
20267 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
20268 Use GET_MODE_CLASS for scalar floating point.
20270 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
20274 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
20275 Fix call of call_cgraph_insertion_hooks.
20277 2018-01-18 Martin Sebor <msebor@redhat.com>
20279 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
20281 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
20284 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
20287 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
20290 * common.opt: (-ffile-prefix-map): New option.
20291 * opts.c (common_handle_option): Defer it.
20292 * opts-global.c (handle_common_deferred_options): Handle it.
20293 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
20294 * file-prefix-map.h: New file.
20295 (remap_debug_filename, add_debug_prefix_map): ...here.
20296 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
20297 * final.c (debug_prefix_map, add_debug_prefix_map
20298 remap_debug_filename): Move to...
20299 * file-prefix-map.c: New file.
20300 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
20301 generalize, get rid of alloca(), use strrchr() instead of strchr().
20302 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
20303 Implement in terms of add_prefix_map().
20304 (remap_macro_filename, remap_debug_filename): Implement in term of
20306 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
20307 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
20308 * dbxout.c: Include file-prefix-map.h.
20309 * varasm.c: Likewise.
20310 * vmsdbgout.c: Likewise.
20311 * xcoffout.c: Likewise.
20312 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
20313 * doc/cppopts.texi (-fmacro-prefix-map): Document.
20314 * doc/invoke.texi (-ffile-prefix-map): Document.
20315 (-fdebug-prefix-map): Update description.
20317 2018-01-18 Martin Liska <mliska@suse.cz>
20319 * config/i386/i386.c (indirect_thunk_name): Document that also
20321 (output_indirect_thunk): Document why both instructions
20322 (pause and lfence) are generated.
20324 2018-01-18 Richard Biener <rguenther@suse.de>
20326 PR tree-optimization/83887
20327 * graphite-scop-detection.c
20328 (scop_detection::get_nearest_dom_with_single_entry): Remove.
20329 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
20330 (scop_detection::merge_sese): Re-implement with a flood-fill
20331 algorithm that properly finds a SESE region if it exists.
20333 2018-01-18 Jakub Jelinek <jakub@redhat.com>
20336 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
20337 pointer_diff optimizations use view_convert instead of convert.
20339 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20341 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20342 Generate different code for -mno-speculate-indirect-jumps.
20343 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20344 (*call_indirect_aix<mode>): Disable for
20345 -mno-speculate-indirect-jumps.
20346 (*call_indirect_aix<mode>_nospec): New define_insn.
20347 (*call_value_indirect_aix<mode>): Disable for
20348 -mno-speculate-indirect-jumps.
20349 (*call_value_indirect_aix<mode>_nospec): New define_insn.
20350 (*sibcall_nonlocal_sysv<mode>): Generate different code for
20351 -mno-speculate-indirect-jumps.
20352 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20354 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
20356 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
20357 long double type, set the flags for noting the default long double
20358 type, even if we don't pass or return a long double type.
20360 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
20363 * ipa-inline.c (flatten_function): Do not overwrite final inlining
20366 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
20368 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
20369 support for merge[hl].
20370 (fold_mergehl_helper): New helper function.
20371 (tree-vector-builder.h): New #include for tree_vector_builder usage.
20372 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
20373 (altivec_vmrglw_direct): Add xxmrglw insn.
20375 2018-01-17 Andrew Waterman <andrew@sifive.com>
20377 * config/riscv/riscv.c (riscv_conditional_register_usage): If
20378 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
20380 2018-01-17 David Malcolm <dmalcolm@redhat.com>
20383 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
20384 call the lto_location_cache before reading the
20385 DECL_SOURCE_LOCATION of the types.
20387 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
20388 Richard Sandiford <richard.sandiford@linaro.org>
20390 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
20391 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
20392 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
20393 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
20394 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
20396 * config/aarch64/constraints.md (aarch64_movti_operand):
20398 * config/aarch64/predicates.md (Uti): Add new constraint.
20400 2018-01-17 Carl Love <cel@us.ibm.com>
20402 * config/rs6000/vsx.md (define_expand xl_len_r,
20403 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
20404 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
20406 (define_expand, define_insn): Move the shift left from the
20407 define_insn to the define_expand for lxvl and stxvl instructions.
20408 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
20409 and XL_LEN_R definitions to PURE.
20411 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
20413 * config/i386/i386.c (indirect_thunk_name): Declare regno
20414 as unsigned int. Compare regno with INVALID_REGNUM.
20415 (output_indirect_thunk): Ditto.
20416 (output_indirect_thunk_function): Ditto.
20417 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
20418 in the call to output_indirect_thunk_function.
20420 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
20422 PR middle-end/83884
20423 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
20424 rather than the size of inner_type to determine the stack slot size
20425 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
20427 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
20430 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
20433 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
20435 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
20436 endian Linux systems to optionally enable multilibs for selecting
20437 the long double type if the user configured an explicit type.
20438 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
20439 have no long double multilibs if not defined.
20440 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
20441 warn if the user used -mabi={ieee,ibm}longdouble and we built
20442 multilibs for long double.
20443 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
20444 appropriate multilib option.
20445 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
20447 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
20448 for building long double multilibs.
20449 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
20451 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
20453 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
20456 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
20458 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
20461 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
20464 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
20467 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
20469 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
20470 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
20472 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
20474 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
20475 different rtl trees depending on TARGET_64BIT.
20476 (rs6000_gen_lvx): Likewise.
20478 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
20480 * config/visium/visium.md (nop): Tweak comment.
20481 (hazard_nop): Likewise.
20483 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20485 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
20486 -mspeculate-indirect-jumps.
20487 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
20488 for -mno-speculate-indirect-jumps.
20489 (*call_indirect_elfv2<mode>_nospec): New define_insn.
20490 (*call_value_indirect_elfv2<mode>): Disable for
20491 -mno-speculate-indirect-jumps.
20492 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
20493 (indirect_jump): Emit different RTL for
20494 -mno-speculate-indirect-jumps.
20495 (*indirect_jump<mode>): Disable for
20496 -mno-speculate-indirect-jumps.
20497 (*indirect_jump<mode>_nospec): New define_insn.
20498 (tablejump): Emit different RTL for
20499 -mno-speculate-indirect-jumps.
20500 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
20501 (tablejumpsi_nospec): New define_expand.
20502 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
20503 (tablejumpdi_nospec): New define_expand.
20504 (*tablejump<mode>_internal1): Disable for
20505 -mno-speculate-indirect-jumps.
20506 (*tablejump<mode>_internal1_nospec): New define_insn.
20507 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
20510 2018-01-16 Artyom Skrobov tyomitch@gmail.com
20512 * caller-save.c (insert_save): Drop unnecessary parameter. All
20515 2018-01-16 Jakub Jelinek <jakub@redhat.com>
20516 Richard Biener <rguenth@suse.de>
20519 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
20520 return early, inline manually is_gimple_sizepos. Make sure if we
20521 call gimplify_expr we don't end up with a gimple constant.
20522 * tree.c (variably_modified_type_p): Don't return true for
20523 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
20524 * gimplify.h (is_gimple_sizepos): Remove.
20526 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
20528 PR tree-optimization/83857
20529 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
20530 vectorizable_live_operation for pure SLP statements.
20531 (vectorizable_live_operation): Handle PHIs.
20533 2018-01-16 Richard Biener <rguenther@suse.de>
20535 PR tree-optimization/83867
20536 * tree-vect-stmts.c (vect_transform_stmt): Precompute
20537 nested_in_vect_loop_p since the scalar stmt may get invalidated.
20539 2018-01-16 Jakub Jelinek <jakub@redhat.com>
20542 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
20543 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
20544 If off is not INTEGER_CST, issue a may not be aligned warning
20545 rather than isn't aligned. Use isn%'t rather than isn't.
20546 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
20548 <case MULT_EXPR>: Improve the case when bottom and one of the
20549 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
20550 operand, in that case check if the other operand is multiple of
20551 bottom divided by the INTEGER_CST operand.
20553 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
20556 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
20557 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
20558 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
20559 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
20560 * config/pa/pa.c (pa_function_arg_advance): Likewise.
20561 (pa_function_arg, pa_arg_partial_bytes): Likewise.
20562 (pa_function_arg_size): New function.
20564 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
20566 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
20567 in a separate statement.
20569 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
20571 PR tree-optimization/83847
20572 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
20573 group gathers and scatters.
20575 2018-01-16 Jakub Jelinek <jakub@redhat.com>
20577 PR rtl-optimization/86620
20578 * params.def (max-sched-ready-insns): Bump minimum value to 1.
20580 PR rtl-optimization/83213
20581 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
20582 to last if both are JUMP_INSNs.
20584 PR tree-optimization/83843
20585 * gimple-ssa-store-merging.c
20586 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
20587 store_immediate_info for bswap/nop orig_stores.
20589 2018-01-15 Andrew Waterman <andrew@sifive.com>
20591 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
20593 <UDIV>: Increase cost if !TARGET_DIV.
20595 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
20597 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
20598 (define_attr "cr_logical_3op"): New.
20599 (cceq_ior_compare): Adjust.
20600 (cceq_ior_compare_complement): Adjust.
20601 (*cceq_rev_compare): Adjust.
20602 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
20603 (is_cracked_insn): Adjust.
20604 (insn_must_be_first_in_group): Adjust.
20605 * config/rs6000/40x.md: Adjust.
20606 * config/rs6000/440.md: Adjust.
20607 * config/rs6000/476.md: Adjust.
20608 * config/rs6000/601.md: Adjust.
20609 * config/rs6000/603.md: Adjust.
20610 * config/rs6000/6xx.md: Adjust.
20611 * config/rs6000/7450.md: Adjust.
20612 * config/rs6000/7xx.md: Adjust.
20613 * config/rs6000/8540.md: Adjust.
20614 * config/rs6000/cell.md: Adjust.
20615 * config/rs6000/e300c2c3.md: Adjust.
20616 * config/rs6000/e500mc.md: Adjust.
20617 * config/rs6000/e500mc64.md: Adjust.
20618 * config/rs6000/e5500.md: Adjust.
20619 * config/rs6000/e6500.md: Adjust.
20620 * config/rs6000/mpc.md: Adjust.
20621 * config/rs6000/power4.md: Adjust.
20622 * config/rs6000/power5.md: Adjust.
20623 * config/rs6000/power6.md: Adjust.
20624 * config/rs6000/power7.md: Adjust.
20625 * config/rs6000/power8.md: Adjust.
20626 * config/rs6000/power9.md: Adjust.
20627 * config/rs6000/rs64.md: Adjust.
20628 * config/rs6000/titan.md: Adjust.
20630 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20632 * config/i386/predicates.md (indirect_branch_operand): Rewrite
20633 ix86_indirect_branch_register logic.
20635 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20637 * config/i386/constraints.md (Bs): Update
20638 ix86_indirect_branch_register check. Don't check
20639 ix86_indirect_branch_register with GOT_memory_operand.
20641 * config/i386/predicates.md (GOT_memory_operand): Don't check
20642 ix86_indirect_branch_register here.
20643 (GOT32_symbol_operand): Likewise.
20645 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20647 * config/i386/predicates.md (constant_call_address_operand):
20648 Rewrite ix86_indirect_branch_register logic.
20649 (sibcall_insn_operand): Likewise.
20651 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20653 * config/i386/constraints.md (Bs): Replace
20654 ix86_indirect_branch_thunk_register with
20655 ix86_indirect_branch_register.
20657 * config/i386/i386.md (indirect_jump): Likewise.
20658 (tablejump): Likewise.
20659 (*sibcall_memory): Likewise.
20660 (*sibcall_value_memory): Likewise.
20661 Peepholes of indirect call and jump via memory: Likewise.
20662 * config/i386/i386.opt: Likewise.
20663 * config/i386/predicates.md (indirect_branch_operand): Likewise.
20664 (GOT_memory_operand): Likewise.
20665 (call_insn_operand): Likewise.
20666 (sibcall_insn_operand): Likewise.
20667 (GOT32_symbol_operand): Likewise.
20669 2018-01-15 Jakub Jelinek <jakub@redhat.com>
20671 PR middle-end/83837
20672 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
20673 type rather than type addr's type points to.
20674 (expand_omp_atomic_mutex): Likewise.
20675 (expand_omp_atomic): Likewise.
20677 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20680 * config/i386/i386.c (output_indirect_thunk_function): Use
20681 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
20682 for __x86_return_thunk.
20684 2018-01-15 Richard Biener <rguenther@suse.de>
20686 PR middle-end/83850
20687 * expmed.c (extract_bit_field_1): Fix typo.
20689 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20692 * config/arm/iterators.md (VF): New mode iterator.
20693 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
20694 Remove integer-related logic from pattern.
20695 (neon_vabd<mode>_3): Likewise.
20697 2018-01-15 Jakub Jelinek <jakub@redhat.com>
20699 PR middle-end/82694
20700 * common.opt (fstrict-overflow): No longer an alias.
20701 (fwrapv-pointer): New option.
20702 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
20703 also for pointer types based on flag_wrapv_pointer.
20704 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
20705 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
20706 opts->x_flag_wrapv got set.
20707 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
20708 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
20709 POINTER_TYPE_OVERFLOW_UNDEFINED.
20710 * match.pd: Likewise in address comparison pattern.
20711 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
20713 2018-01-15 Richard Biener <rguenther@suse.de>
20716 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
20717 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
20718 Reset type names to their identifier if their TYPE_DECL doesn't
20719 have linkage (and thus is used for ODR and devirt).
20720 (save_debug_info_for_decl): Remove.
20721 (save_debug_info_for_type): Likewise.
20722 (add_tree_to_fld_list): Adjust.
20723 * tree-pretty-print.c (dump_generic_node): Make dumping of
20724 type names more robust.
20726 2018-01-15 Richard Biener <rguenther@suse.de>
20728 * BASE-VER: Bump to 8.0.1.
20730 2018-01-14 Martin Sebor <msebor@redhat.com>
20733 * builtins.c (check_access): Avoid warning when the no-warning bit
20736 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
20738 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
20739 * ira-color (allocno_hard_regs_compare): Likewise.
20741 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
20744 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
20745 Use .pushsection/.popsection.
20747 2018-01-14 Martin Sebor <msebor@redhat.com>
20750 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
20752 2018-01-14 Jakub Jelinek <jakub@redhat.com>
20754 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
20755 entry from extra_headers.
20756 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
20757 extra_headers, make the list bitwise identical to the i?86-*-* one.
20759 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20761 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
20762 -mcmodel=large with -mindirect-branch=thunk,
20763 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
20764 -mfunction-return=thunk-extern.
20765 * doc/invoke.texi: Document -mcmodel=large is incompatible with
20766 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
20767 -mfunction-return=thunk and -mfunction-return=thunk-extern.
20769 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20771 * config/i386/i386.c (print_reg): Print the name of the full
20772 integer register without '%'.
20773 (ix86_print_operand): Handle 'V'.
20774 * doc/extend.texi: Document 'V' modifier.
20776 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20778 * config/i386/constraints.md (Bs): Disallow memory operand for
20779 -mindirect-branch-register.
20781 * config/i386/predicates.md (indirect_branch_operand): Likewise.
20782 (GOT_memory_operand): Likewise.
20783 (call_insn_operand): Likewise.
20784 (sibcall_insn_operand): Likewise.
20785 (GOT32_symbol_operand): Likewise.
20786 * config/i386/i386.md (indirect_jump): Call convert_memory_address
20787 for -mindirect-branch-register.
20788 (tablejump): Likewise.
20789 (*sibcall_memory): Likewise.
20790 (*sibcall_value_memory): Likewise.
20791 Disallow peepholes of indirect call and jump via memory for
20792 -mindirect-branch-register.
20793 (*call_pop): Replace m with Bw.
20794 (*call_value_pop): Likewise.
20795 (*sibcall_pop_memory): Replace m with Bs.
20796 * config/i386/i386.opt (mindirect-branch-register): New option.
20797 * doc/invoke.texi: Document -mindirect-branch-register option.
20799 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20801 * config/i386/i386-protos.h (ix86_output_function_return): New.
20802 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
20803 set function_return_type.
20804 (indirect_thunk_name): Add ret_p to indicate thunk for function
20806 (output_indirect_thunk_function): Pass false to
20807 indirect_thunk_name.
20808 (ix86_output_indirect_branch_via_reg): Likewise.
20809 (ix86_output_indirect_branch_via_push): Likewise.
20810 (output_indirect_thunk_function): Create alias for function
20811 return thunk if regno < 0.
20812 (ix86_output_function_return): New function.
20813 (ix86_handle_fndecl_attribute): Handle function_return.
20814 (ix86_attribute_table): Add function_return.
20815 * config/i386/i386.h (machine_function): Add
20816 function_return_type.
20817 * config/i386/i386.md (simple_return_internal): Use
20818 ix86_output_function_return.
20819 (simple_return_internal_long): Likewise.
20820 * config/i386/i386.opt (mfunction-return=): New option.
20821 (indirect_branch): Mention -mfunction-return=.
20822 * doc/extend.texi: Document function_return function attribute.
20823 * doc/invoke.texi: Document -mfunction-return= option.
20825 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20827 * config/i386/i386-opts.h (indirect_branch): New.
20828 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
20829 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
20830 with local indirect jump when converting indirect call and jump.
20831 (ix86_set_indirect_branch_type): New.
20832 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
20833 (indirectlabelno): New.
20834 (indirect_thunk_needed): Likewise.
20835 (indirect_thunk_bnd_needed): Likewise.
20836 (indirect_thunks_used): Likewise.
20837 (indirect_thunks_bnd_used): Likewise.
20838 (INDIRECT_LABEL): Likewise.
20839 (indirect_thunk_name): Likewise.
20840 (output_indirect_thunk): Likewise.
20841 (output_indirect_thunk_function): Likewise.
20842 (ix86_output_indirect_branch_via_reg): Likewise.
20843 (ix86_output_indirect_branch_via_push): Likewise.
20844 (ix86_output_indirect_branch): Likewise.
20845 (ix86_output_indirect_jmp): Likewise.
20846 (ix86_code_end): Call output_indirect_thunk_function if needed.
20847 (ix86_output_call_insn): Call ix86_output_indirect_branch if
20849 (ix86_handle_fndecl_attribute): Handle indirect_branch.
20850 (ix86_attribute_table): Add indirect_branch.
20851 * config/i386/i386.h (machine_function): Add indirect_branch_type
20852 and has_local_indirect_jump.
20853 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
20855 (tablejump): Likewise.
20856 (*indirect_jump): Use ix86_output_indirect_jmp.
20857 (*tablejump_1): Likewise.
20858 (simple_return_indirect_internal): Likewise.
20859 * config/i386/i386.opt (mindirect-branch=): New option.
20860 (indirect_branch): New.
20863 (thunk-inline): Likewise.
20864 (thunk-extern): Likewise.
20865 * doc/extend.texi: Document indirect_branch function attribute.
20866 * doc/invoke.texi: Document -mindirect-branch= option.
20868 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
20871 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
20873 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
20875 * ipa-inline.c (want_inline_small_function_p): Return false if
20876 inlining has already failed with CIF_FINAL_ERROR.
20877 (update_caller_keys): Call want_inline_small_function_p before
20879 (update_callee_keys): Likewise.
20881 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
20883 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
20885 (rs6000_quadword_masked_address_p): Likewise.
20886 (quad_aligned_load_p): Likewise.
20887 (quad_aligned_store_p): Likewise.
20888 (const_load_sequence_p): Add comment to describe the outer-most loop.
20889 (mimic_memory_attributes_and_flags): New function.
20890 (rs6000_gen_stvx): Likewise.
20891 (replace_swapped_aligned_store): Likewise.
20892 (rs6000_gen_lvx): Likewise.
20893 (replace_swapped_aligned_load): Likewise.
20894 (replace_swapped_load_constant): Capitalize argument name in
20895 comment describing this function.
20896 (rs6000_analyze_swaps): Add a third pass to search for vector loads
20897 and stores that access quad-word aligned addresses and replace
20898 with stvx or lvx instructions when appropriate.
20899 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
20900 New function prototype.
20901 (rs6000_quadword_masked_address_p): Likewise.
20902 (rs6000_gen_lvx): Likewise.
20903 (rs6000_gen_stvx): Likewise.
20904 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
20905 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
20906 when memory address is aligned.
20907 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
20908 this split to select lvx instruction when memory address is aligned.
20909 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
20910 instruction when memory address is aligned.
20911 (*vsx_le_perm_load_v16qi): Likewise.
20912 (four unnamed splitters): Modify to select the stvx instruction
20913 when memory is aligned.
20915 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
20917 * predict.c (determine_unlikely_bbs): Handle correctly BBs
20918 which appears in the queue multiple times.
20920 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20921 Alan Hayward <alan.hayward@arm.com>
20922 David Sherwood <david.sherwood@arm.com>
20924 * tree-vectorizer.h (vec_lower_bound): New structure.
20925 (_loop_vec_info): Add check_nonzero and lower_bounds.
20926 (LOOP_VINFO_CHECK_NONZERO): New macro.
20927 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
20928 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
20929 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
20930 fields. Make seg_len the distance travelled, not including the
20932 (dr_direction_indicator): Declare.
20933 (dr_zero_step_indicator): Likewise.
20934 (dr_known_forward_stride_p): Likewise.
20935 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
20937 (runtime_alias_check_p): Allow runtime alias checks with
20939 (operator ==): Compare access_size and align.
20940 (prune_runtime_alias_test_list): Rework for new distinction between
20941 the access_size and seg_len.
20942 (create_intersect_range_checks_index): Likewise. Cope with polynomial
20944 (get_segment_min_max): New function.
20945 (create_intersect_range_checks): Use it.
20946 (dr_step_indicator): New function.
20947 (dr_direction_indicator): Likewise.
20948 (dr_zero_step_indicator): Likewise.
20949 (dr_known_forward_stride_p): Likewise.
20950 * tree-loop-distribution.c (data_ref_segment_size): Return
20951 DR_STEP * (niters - 1).
20952 (compute_alias_check_pairs): Update call to the dr_with_seg_len
20954 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
20955 (vect_preserves_scalar_order_p): New function, split out from...
20956 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
20957 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
20958 (vect_vfa_access_size): New function.
20959 (vect_vfa_align): Likewise.
20960 (vect_compile_time_alias): Take access_size_a and access_b arguments.
20961 (dump_lower_bound): New function.
20962 (vect_check_lower_bound): Likewise.
20963 (vect_small_gap_p): Likewise.
20964 (vectorizable_with_step_bound_p): Likewise.
20965 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
20966 depencies if the vectorization factor is 1. Convert the checks
20967 for nonzero steps into checks on the bounds of DR_STEP. Try using
20968 a bunds check for variable steps if the minimum required step is
20969 relatively small. Update calls to the dr_with_seg_len
20970 constructor and to vect_compile_time_alias.
20971 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
20973 (vect_loop_versioning): Call it.
20974 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
20976 (vect_estimate_min_profitable_iters): Account for any bounds checks.
20978 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20979 Alan Hayward <alan.hayward@arm.com>
20980 David Sherwood <david.sherwood@arm.com>
20982 * doc/sourcebuild.texi (vect_scatter_store): Document.
20983 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
20985 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
20987 * genopinit.c (main): Add supports_vec_scatter_store and
20988 supports_vec_scatter_store_cached to target_optabs.
20989 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
20990 IFN_MASK_SCATTER_STORE.
20991 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
20993 * internal-fn.h (internal_store_fn_p): Declare.
20994 (internal_fn_stored_value_index): Likewise.
20995 * internal-fn.c (scatter_store_direct): New macro.
20996 (expand_scatter_store_optab_fn): New function.
20997 (direct_scatter_store_optab_supported_p): New macro.
20998 (internal_store_fn_p): New function.
20999 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
21000 IFN_MASK_SCATTER_STORE.
21001 (internal_fn_mask_index): Likewise.
21002 (internal_fn_stored_value_index): New function.
21003 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
21004 for scatter stores.
21005 * optabs-query.h (supports_vec_scatter_store_p): Declare.
21006 * optabs-query.c (supports_vec_scatter_store_p): New function.
21007 * tree-vectorizer.h (vect_get_store_rhs): Declare.
21008 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
21009 true for scatter stores.
21010 (vect_gather_scatter_fn_p): Handle scatter stores too.
21011 (vect_check_gather_scatter): Consider using scatter stores if
21012 supports_vec_scatter_store_p.
21013 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
21014 scatter stores too.
21015 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
21016 internal_fn_stored_value_index.
21017 (check_load_store_masking): Handle scatter stores too.
21018 (vect_get_store_rhs): Make public.
21019 (vectorizable_call): Use internal_store_fn_p.
21020 (vectorizable_store): Handle scatter store internal functions.
21021 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
21022 when deciding whether the end of the group has been reached.
21023 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
21024 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
21025 (mask_scatter_store<mode>): New insns.
21027 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21028 Alan Hayward <alan.hayward@arm.com>
21029 David Sherwood <david.sherwood@arm.com>
21031 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
21032 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
21033 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
21035 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
21036 Use vect_truncate_gather_scatter_offset if we can't treat the
21037 operation as a normal gather load or scatter store.
21038 (get_group_load_store_type): Take the gather_scatter_info
21039 as argument. Try using a gather load or scatter store for
21040 single-element groups.
21041 (get_load_store_type): Update calls to get_group_load_store_type
21042 and vect_use_strided_gather_scatters_p.
21044 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21045 Alan Hayward <alan.hayward@arm.com>
21046 David Sherwood <david.sherwood@arm.com>
21048 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
21049 optional tree argument.
21050 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
21052 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
21053 but continue to use the current value as a fallback.
21054 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
21055 to compare the updates.
21056 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
21057 (get_load_store_type): Use it when handling a strided access.
21058 (vect_get_strided_load_store_ops): New function.
21059 (vect_get_data_ptr_increment): Likewise.
21060 (vectorizable_load): Handle strided gather loads. Always pass
21061 a step to vect_create_data_ref_ptr and bump_vector_ptr.
21063 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21064 Alan Hayward <alan.hayward@arm.com>
21065 David Sherwood <david.sherwood@arm.com>
21067 * doc/md.texi (gather_load@var{m}): Document.
21068 (mask_gather_load@var{m}): Likewise.
21069 * genopinit.c (main): Add supports_vec_gather_load and
21070 supports_vec_gather_load_cached to target_optabs.
21071 * optabs-tree.c (init_tree_optimization_optabs): Use
21072 ggc_cleared_alloc to allocate target_optabs.
21073 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
21074 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
21076 * internal-fn.h (internal_load_fn_p): Declare.
21077 (internal_gather_scatter_fn_p): Likewise.
21078 (internal_fn_mask_index): Likewise.
21079 (internal_gather_scatter_fn_supported_p): Likewise.
21080 * internal-fn.c (gather_load_direct): New macro.
21081 (expand_gather_load_optab_fn): New function.
21082 (direct_gather_load_optab_supported_p): New macro.
21083 (direct_internal_fn_optab): New function.
21084 (internal_load_fn_p): Likewise.
21085 (internal_gather_scatter_fn_p): Likewise.
21086 (internal_fn_mask_index): Likewise.
21087 (internal_gather_scatter_fn_supported_p): Likewise.
21088 * optabs-query.c (supports_at_least_one_mode_p): New function.
21089 (supports_vec_gather_load_p): Likewise.
21090 * optabs-query.h (supports_vec_gather_load_p): Declare.
21091 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
21092 and memory_type field.
21093 (NUM_PATTERNS): Bump to 15.
21094 * tree-vect-data-refs.c: Include internal-fn.h.
21095 (vect_gather_scatter_fn_p): New function.
21096 (vect_describe_gather_scatter_call): Likewise.
21097 (vect_check_gather_scatter): Try using internal functions for
21098 gather loads. Recognize existing calls to a gather load function.
21099 (vect_analyze_data_refs): Consider using gather loads if
21100 supports_vec_gather_load_p.
21101 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
21102 (vect_get_gather_scatter_offset_type): Likewise.
21103 (vect_convert_mask_for_vectype): Likewise.
21104 (vect_add_conversion_to_patterm): Likewise.
21105 (vect_try_gather_scatter_pattern): Likewise.
21106 (vect_recog_gather_scatter_pattern): New pattern recognizer.
21107 (vect_vect_recog_func_ptrs): Add it.
21108 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
21109 internal_fn_mask_index and internal_gather_scatter_fn_p.
21110 (check_load_store_masking): Take the gather_scatter_info as an
21111 argument and handle gather loads.
21112 (vect_get_gather_scatter_ops): New function.
21113 (vectorizable_call): Check internal_load_fn_p.
21114 (vectorizable_load): Likewise. Handle gather load internal
21116 (vectorizable_store): Update call to check_load_store_masking.
21117 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
21118 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
21119 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
21120 (aarch64_gather_scale_operand_d): New predicates.
21121 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
21122 (mask_gather_load<mode>): New insns.
21124 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21125 Alan Hayward <alan.hayward@arm.com>
21126 David Sherwood <david.sherwood@arm.com>
21128 * optabs.def (fold_left_plus_optab): New optab.
21129 * doc/md.texi (fold_left_plus_@var{m}): Document.
21130 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
21131 * internal-fn.c (fold_left_direct): Define.
21132 (expand_fold_left_optab_fn): Likewise.
21133 (direct_fold_left_optab_supported_p): Likewise.
21134 * fold-const-call.c (fold_const_fold_left): New function.
21135 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
21136 * tree-parloops.c (valid_reduction_p): New function.
21137 (gather_scalar_reductions): Use it.
21138 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
21139 (vect_finish_replace_stmt): Declare.
21140 * tree-vect-loop.c (fold_left_reduction_fn): New function.
21141 (needs_fold_left_reduction_p): New function, split out from...
21142 (vect_is_simple_reduction): ...here. Accept reductions that
21143 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
21144 (vect_force_simple_reduction): Also store the reduction type in
21145 the assignment's STMT_VINFO_REDUC_TYPE.
21146 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
21147 (merge_with_identity): New function.
21148 (vect_expand_fold_left): Likewise.
21149 (vectorize_fold_left_reduction): Likewise.
21150 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
21151 scalar phi in place for it. Check for target support and reject
21152 cases that would reassociate the operation. Defer the transform
21153 phase to vectorize_fold_left_reduction.
21154 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
21155 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
21156 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
21158 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21160 * tree-if-conv.c (predicate_mem_writes): Remove redundant
21161 call to ifc_temp_var.
21163 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21164 Alan Hayward <alan.hayward@arm.com>
21165 David Sherwood <david.sherwood@arm.com>
21167 * target.def (legitimize_address_displacement): Take the original
21168 offset as a poly_int.
21169 * targhooks.h (default_legitimize_address_displacement): Update
21171 * targhooks.c (default_legitimize_address_displacement): Likewise.
21172 * doc/tm.texi: Regenerate.
21173 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
21174 as an argument, moving assert of ad->disp == ad->disp_term to...
21175 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
21176 Try calling targetm.legitimize_address_displacement before expanding
21177 the address rather than afterwards, and adjust for the new interface.
21178 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
21179 Match the new hook interface. Handle SVE addresses.
21180 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
21181 new hook interface.
21183 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21185 * Makefile.in (OBJS): Add early-remat.o.
21186 * target.def (select_early_remat_modes): New hook.
21187 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
21188 * doc/tm.texi: Regenerate.
21189 * targhooks.h (default_select_early_remat_modes): Declare.
21190 * targhooks.c (default_select_early_remat_modes): New function.
21191 * timevar.def (TV_EARLY_REMAT): New timevar.
21192 * passes.def (pass_early_remat): New pass.
21193 * tree-pass.h (make_pass_early_remat): Declare.
21194 * early-remat.c: New file.
21195 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
21197 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
21199 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21200 Alan Hayward <alan.hayward@arm.com>
21201 David Sherwood <david.sherwood@arm.com>
21203 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
21204 vfm1 with a bound_epilog parameter.
21205 (vect_do_peeling): Update calls accordingly, and move the prologue
21206 call earlier in the function. Treat the base bound_epilog as 0 for
21207 fully-masked loops and retain vf - 1 for other loops. Add 1 to
21208 this base when peeling for gaps.
21209 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
21210 with fully-masked loops.
21211 (vect_estimate_min_profitable_iters): Handle the single peeled
21212 iteration in that case.
21214 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21215 Alan Hayward <alan.hayward@arm.com>
21216 David Sherwood <david.sherwood@arm.com>
21218 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
21219 single-element interleaving even if the size is not a power of 2.
21220 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
21221 accesses for single-element interleaving if the group size is
21224 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21225 Alan Hayward <alan.hayward@arm.com>
21226 David Sherwood <david.sherwood@arm.com>
21228 * doc/md.texi (fold_extract_last_@var{m}): Document.
21229 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
21230 * optabs.def (fold_extract_last_optab): New optab.
21231 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
21232 * internal-fn.c (fold_extract_direct): New macro.
21233 (expand_fold_extract_optab_fn): Likewise.
21234 (direct_fold_extract_optab_supported_p): Likewise.
21235 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
21236 * tree-vect-loop.c (vect_model_reduction_cost): Handle
21237 EXTRACT_LAST_REDUCTION.
21238 (get_initial_def_for_reduction): Do not create an initial vector
21239 for EXTRACT_LAST_REDUCTION reductions.
21240 (vectorizable_reduction): Leave the scalar phi in place for
21241 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
21242 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
21243 epilogue code for EXTRACT_LAST_REDUCTION and defer the
21244 transform phase to vectorizable_condition.
21245 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
21247 (vect_finish_stmt_generation): ...here.
21248 (vect_finish_replace_stmt): New function.
21249 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
21250 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
21252 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
21254 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21255 Alan Hayward <alan.hayward@arm.com>
21256 David Sherwood <david.sherwood@arm.com>
21258 * doc/md.texi (extract_last_@var{m}): Document.
21259 * optabs.def (extract_last_optab): New optab.
21260 * internal-fn.def (EXTRACT_LAST): New internal function.
21261 * internal-fn.c (cond_unary_direct): New macro.
21262 (expand_cond_unary_optab_fn): Likewise.
21263 (direct_cond_unary_optab_supported_p): Likewise.
21264 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
21265 loops using EXTRACT_LAST.
21266 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
21267 (extract_last_<mode>): ...this optab.
21268 (vec_extract<mode><Vel>): Update accordingly.
21270 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21271 Alan Hayward <alan.hayward@arm.com>
21272 David Sherwood <david.sherwood@arm.com>
21274 * target.def (empty_mask_is_expensive): New hook.
21275 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
21276 * doc/tm.texi: Regenerate.
21277 * targhooks.h (default_empty_mask_is_expensive): Declare.
21278 * targhooks.c (default_empty_mask_is_expensive): New function.
21279 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
21280 if the target says that empty masks are expensive.
21281 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
21283 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
21285 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21286 Alan Hayward <alan.hayward@arm.com>
21287 David Sherwood <david.sherwood@arm.com>
21289 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
21290 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
21291 (vect_use_loop_mask_for_alignment_p): New function.
21292 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
21293 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
21294 niters_skip argument. Make sure that the first niters_skip elements
21295 of the first iteration are inactive.
21296 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
21297 Update call to vect_set_loop_masks_directly.
21298 (get_misalign_in_elems): New function, split out from...
21299 (vect_gen_prolog_loop_niters): ...here.
21300 (vect_update_init_of_dr): Take a code argument that specifies whether
21301 the adjustment should be added or subtracted.
21302 (vect_update_init_of_drs): Likewise.
21303 (vect_prepare_for_masked_peels): New function.
21304 (vect_do_peeling): Skip prologue peeling if we're using a mask
21305 instead. Update call to vect_update_inits_of_drs.
21306 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21308 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
21309 alignment. Do not include the number of peeled iterations in
21310 the minimum threshold in that case.
21311 (vectorizable_induction): Adjust the start value down by
21312 LOOP_VINFO_MASK_SKIP_NITERS iterations.
21313 (vect_transform_loop): Call vect_prepare_for_masked_peels.
21314 Take the number of skipped iterations into account when calculating
21316 * tree-vect-stmts.c (vect_gen_while_not): New function.
21318 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21319 Alan Hayward <alan.hayward@arm.com>
21320 David Sherwood <david.sherwood@arm.com>
21322 * doc/sourcebuild.texi (vect_fully_masked): Document.
21323 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
21324 default value to 0.
21325 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
21327 (vect_analyze_loop_2): ...here. Don't check the vectorization
21328 factor against the number of loop iterations if the loop is
21331 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21332 Alan Hayward <alan.hayward@arm.com>
21333 David Sherwood <david.sherwood@arm.com>
21335 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
21336 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
21337 (dump_groups): Update accordingly.
21338 (iv_use::mem_type): New member variable.
21339 (address_p): New function.
21340 (record_use): Add a mem_type argument and initialize the new
21342 (record_group_use): Add a mem_type argument. Use address_p.
21343 Remove obsolete null checks of base_object. Update call to record_use.
21344 (find_interesting_uses_op): Update call to record_group_use.
21345 (find_interesting_uses_cond): Likewise.
21346 (find_interesting_uses_address): Likewise.
21347 (get_mem_type_for_internal_fn): New function.
21348 (find_address_like_use): Likewise.
21349 (find_interesting_uses_stmt): Try find_address_like_use before
21350 calling find_interesting_uses_op.
21351 (addr_offset_valid_p): Use the iv mem_type field as the type
21352 of the addressed memory.
21353 (add_autoinc_candidates): Likewise.
21354 (get_address_cost): Likewise.
21355 (split_small_address_groups_p): Use address_p.
21356 (split_address_groups): Likewise.
21357 (add_iv_candidate_for_use): Likewise.
21358 (autoinc_possible_for_pair): Likewise.
21359 (rewrite_groups): Likewise.
21360 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
21361 (determine_group_iv_cost): Update after split of USE_ADDRESS.
21362 (get_alias_ptr_type_for_ptr_address): New function.
21363 (rewrite_use_address): Rewrite address uses in calls that were
21364 identified by find_address_like_use.
21366 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21367 Alan Hayward <alan.hayward@arm.com>
21368 David Sherwood <david.sherwood@arm.com>
21370 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
21372 * gimple-expr.h (is_gimple_addressable: Likewise.
21373 * gimple-expr.c (is_gimple_address): Likewise.
21374 * internal-fn.c (expand_call_mem_ref): New function.
21375 (expand_mask_load_optab_fn): Use it.
21376 (expand_mask_store_optab_fn): Likewise.
21378 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21379 Alan Hayward <alan.hayward@arm.com>
21380 David Sherwood <david.sherwood@arm.com>
21382 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
21383 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
21384 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
21385 (cond_umax@var{mode}): Document.
21386 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
21387 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
21388 (cond_umin_optab, cond_umax_optab): New optabs.
21389 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
21390 (COND_IOR, COND_XOR): New internal functions.
21391 * internal-fn.h (get_conditional_internal_fn): Declare.
21392 * internal-fn.c (cond_binary_direct): New macro.
21393 (expand_cond_binary_optab_fn): Likewise.
21394 (direct_cond_binary_optab_supported_p): Likewise.
21395 (get_conditional_internal_fn): New function.
21396 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
21397 Cope with reduction statements that are vectorized as calls rather
21399 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
21400 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
21401 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
21402 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
21403 (UNSPEC_COND_EOR): New unspecs.
21404 (optab): Add mappings for them.
21405 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
21406 (sve_int_op, sve_fp_op): New int attributes.
21408 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21409 Alan Hayward <alan.hayward@arm.com>
21410 David Sherwood <david.sherwood@arm.com>
21412 * optabs.def (while_ult_optab): New optab.
21413 * doc/md.texi (while_ult@var{m}@var{n}): Document.
21414 * internal-fn.def (WHILE_ULT): New internal function.
21415 * internal-fn.h (direct_internal_fn_supported_p): New override
21416 that takes two types as argument.
21417 * internal-fn.c (while_direct): New macro.
21418 (expand_while_optab_fn): New function.
21419 (convert_optab_supported_p): Likewise.
21420 (direct_while_optab_supported_p): New macro.
21421 * wide-int.h (wi::udiv_ceil): New function.
21422 * tree-vectorizer.h (rgroup_masks): New structure.
21423 (vec_loop_masks): New typedef.
21424 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
21425 and fully_masked_p.
21426 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
21427 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
21428 (vect_max_vf): New function.
21429 (slpeel_make_loop_iterate_ntimes): Delete.
21430 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
21431 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
21432 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
21433 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
21434 internal-fn.h, stor-layout.h and optabs-query.h.
21435 (vect_set_loop_mask): New function.
21436 (add_preheader_seq): Likewise.
21437 (add_header_seq): Likewise.
21438 (interleave_supported_p): Likewise.
21439 (vect_maybe_permute_loop_masks): Likewise.
21440 (vect_set_loop_masks_directly): Likewise.
21441 (vect_set_loop_condition_masked): Likewise.
21442 (vect_set_loop_condition_unmasked): New function, split out from
21443 slpeel_make_loop_iterate_ntimes.
21444 (slpeel_make_loop_iterate_ntimes): Rename to..
21445 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
21446 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
21447 (vect_do_peeling): Update call accordingly.
21448 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
21450 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21451 mask_compare_type, can_fully_mask_p and fully_masked_p.
21452 (release_vec_loop_masks): New function.
21453 (_loop_vec_info): Use it to free the loop masks.
21454 (can_produce_all_loop_masks_p): New function.
21455 (vect_get_max_nscalars_per_iter): Likewise.
21456 (vect_verify_full_masking): Likewise.
21457 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
21458 retries, and free the mask rgroups before retrying. Check loop-wide
21459 reasons for disallowing fully-masked loops. Make the final decision
21460 about whether use a fully-masked loop or not.
21461 (vect_estimate_min_profitable_iters): Do not assume that peeling
21462 for the number of iterations will be needed for fully-masked loops.
21463 (vectorizable_reduction): Disable fully-masked loops.
21464 (vectorizable_live_operation): Likewise.
21465 (vect_halve_mask_nunits): New function.
21466 (vect_double_mask_nunits): Likewise.
21467 (vect_record_loop_mask): Likewise.
21468 (vect_get_loop_mask): Likewise.
21469 (vect_transform_loop): Handle the case in which the final loop
21470 iteration might handle a partial vector. Call vect_set_loop_condition
21471 instead of slpeel_make_loop_iterate_ntimes.
21472 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
21473 (check_load_store_masking): New function.
21474 (prepare_load_store_mask): Likewise.
21475 (vectorizable_store): Handle fully-masked loops.
21476 (vectorizable_load): Likewise.
21477 (supportable_widening_operation): Use vect_halve_mask_nunits for
21479 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
21480 (vect_gen_while): New function.
21481 * config/aarch64/aarch64.md (umax<mode>3): New expander.
21482 (aarch64_uqdec<mode>): New insn.
21484 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21485 Alan Hayward <alan.hayward@arm.com>
21486 David Sherwood <david.sherwood@arm.com>
21488 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
21489 (reduc_xor_scal_optab): New optabs.
21490 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
21491 (reduc_xor_scal_@var{m}): Document.
21492 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
21493 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
21494 internal functions.
21495 * fold-const-call.c (fold_const_call): Handle them.
21496 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
21497 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
21498 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
21499 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
21500 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
21501 (UNSPEC_XORV): New unspecs.
21502 (optab): Add entries for them.
21503 (BITWISEV): New int iterator.
21504 (bit_reduc_op): New int attributes.
21506 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21507 Alan Hayward <alan.hayward@arm.com>
21508 David Sherwood <david.sherwood@arm.com>
21510 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
21511 * internal-fn.def (VEC_SHL_INSERT): New internal function.
21512 * optabs.def (vec_shl_insert_optab): New optab.
21513 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
21514 (duplicate_and_interleave): Likewise.
21515 * tree-vect-loop.c: Include internal-fn.h.
21516 (neutral_op_for_slp_reduction): New function, split out from
21517 get_initial_defs_for_reduction.
21518 (get_initial_def_for_reduction): Handle option 2 for variable-length
21519 vectors by loading the neutral value into a vector and then shifting
21520 the initial value into element 0.
21521 (get_initial_defs_for_reduction): Replace the code argument with
21522 the neutral value calculated by neutral_op_for_slp_reduction.
21523 Use gimple_build_vector for constant-length vectors.
21524 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
21525 but the first group_size elements have a neutral value.
21526 Use duplicate_and_interleave otherwise.
21527 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
21528 Update call to get_initial_defs_for_reduction. Handle SLP
21529 reductions for variable-length vectors by creating one vector
21530 result for each scalar result, with the elements associated
21531 with other scalar results stubbed out with the neutral value.
21532 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
21533 Require IFN_VEC_SHL_INSERT for double reductions on
21534 variable-length vectors, or SLP reductions that have
21535 a neutral value. Require can_duplicate_and_interleave_p
21536 support for variable-length unchained SLP reductions if there
21537 is no neutral value, such as for MIN/MAX reductions. Also require
21538 the number of vector elements to be a multiple of the number of
21539 SLP statements when doing variable-length unchained SLP reductions.
21540 Update call to vect_create_epilog_for_reduction.
21541 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
21542 and remove initial values.
21543 (duplicate_and_interleave): Make public.
21544 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
21545 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
21547 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21548 Alan Hayward <alan.hayward@arm.com>
21549 David Sherwood <david.sherwood@arm.com>
21551 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
21552 (can_duplicate_and_interleave_p): New function.
21553 (vect_get_and_check_slp_defs): Take the vector of statements
21554 rather than just the current one. Remove excess parentheses.
21555 Restriction rejectinon of vect_constant_def and vect_external_def
21556 for variable-length vectors to boolean types, or types for which
21557 can_duplicate_and_interleave_p is false.
21558 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
21559 (duplicate_and_interleave): New function.
21560 (vect_get_constant_vectors): Use gimple_build_vector for
21561 constant-length vectors and suitable variable-length constant
21562 vectors. Use duplicate_and_interleave for other variable-length
21563 vectors. Don't defer the update when inserting new statements.
21565 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21566 Alan Hayward <alan.hayward@arm.com>
21567 David Sherwood <david.sherwood@arm.com>
21569 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
21570 min_profitable_iters doesn't go negative.
21572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21573 Alan Hayward <alan.hayward@arm.com>
21574 David Sherwood <david.sherwood@arm.com>
21576 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
21577 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
21578 * optabs.def (vec_mask_load_lanes_optab): New optab.
21579 (vec_mask_store_lanes_optab): Likewise.
21580 * internal-fn.def (MASK_LOAD_LANES): New internal function.
21581 (MASK_STORE_LANES): Likewise.
21582 * internal-fn.c (mask_load_lanes_direct): New macro.
21583 (mask_store_lanes_direct): Likewise.
21584 (expand_mask_load_optab_fn): Handle masked operations.
21585 (expand_mask_load_lanes_optab_fn): New macro.
21586 (expand_mask_store_optab_fn): Handle masked operations.
21587 (expand_mask_store_lanes_optab_fn): New macro.
21588 (direct_mask_load_lanes_optab_supported_p): Likewise.
21589 (direct_mask_store_lanes_optab_supported_p): Likewise.
21590 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
21592 (vect_load_lanes_supported): Likewise.
21593 * tree-vect-data-refs.c (strip_conversion): New function.
21594 (can_group_stmts_p): Likewise.
21595 (vect_analyze_data_ref_accesses): Use it instead of checking
21596 for a pair of assignments.
21597 (vect_store_lanes_supported): Take a masked_p parameter.
21598 (vect_load_lanes_supported): Likewise.
21599 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
21600 vect_store_lanes_supported and vect_load_lanes_supported.
21601 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
21602 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
21603 parameter. Don't allow gaps for masked accesses.
21604 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
21605 and vect_load_lanes_supported.
21606 (get_load_store_type): Take a masked_p parameter and update
21607 call to get_group_load_store_type.
21608 (vectorizable_store): Update call to get_load_store_type.
21609 Handle IFN_MASK_STORE_LANES.
21610 (vectorizable_load): Update call to get_load_store_type.
21611 Handle IFN_MASK_LOAD_LANES.
21613 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21614 Alan Hayward <alan.hayward@arm.com>
21615 David Sherwood <david.sherwood@arm.com>
21617 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
21619 * config/aarch64/aarch64-protos.h
21620 (aarch64_sve_struct_memory_operand_p): Declare.
21621 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
21622 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
21623 (VPRED, vpred): Handle SVE structure modes.
21624 * config/aarch64/constraints.md (Utx): New constraint.
21625 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
21626 (aarch64_sve_struct_nonimmediate_operand): New predicates.
21627 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
21628 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
21629 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
21630 structure modes. Split into pieces after RA.
21631 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
21632 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
21634 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
21635 SVE structure modes.
21636 (aarch64_classify_address): Likewise.
21637 (sizetochar): Move earlier in file.
21638 (aarch64_print_operand): Handle SVE register lists.
21639 (aarch64_array_mode): New function.
21640 (aarch64_sve_struct_memory_operand_p): Likewise.
21641 (TARGET_ARRAY_MODE): Redefine.
21643 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21644 Alan Hayward <alan.hayward@arm.com>
21645 David Sherwood <david.sherwood@arm.com>
21647 * target.def (array_mode): New target hook.
21648 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
21649 * doc/tm.texi: Regenerate.
21650 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
21651 * hooks.c (hook_optmode_mode_uhwi_none): New function.
21652 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
21653 targetm.array_mode.
21654 * stor-layout.c (mode_for_array): Likewise. Support polynomial
21657 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21658 Alan Hayward <alan.hayward@arm.com>
21659 David Sherwood <david.sherwood@arm.com>
21661 * fold-const.c (fold_binary_loc): Check the argument types
21662 rather than the result type when testing for a vector operation.
21664 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21666 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
21667 * doc/tm.texi: Regenerate.
21669 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21670 Alan Hayward <alan.hayward@arm.com>
21671 David Sherwood <david.sherwood@arm.com>
21673 * doc/invoke.texi (-msve-vector-bits=): Document new option.
21674 (sve): Document new AArch64 extension.
21675 * doc/md.texi (w): Extend the description of the AArch64
21676 constraint to include SVE vectors.
21677 (Upl, Upa): Document new AArch64 predicate constraints.
21678 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
21680 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
21681 (msve-vector-bits=): New option.
21682 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
21683 SVE when these are disabled.
21684 (sve): New extension.
21685 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
21686 modes. Adjust their number of units based on aarch64_sve_vg.
21687 (MAX_BITSIZE_MODE_ANY_MODE): Define.
21688 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
21689 aarch64_addr_query_type.
21690 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
21691 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
21692 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
21693 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
21694 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
21695 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
21696 (aarch64_simd_imm_zero_p): Delete.
21697 (aarch64_check_zero_based_sve_index_immediate): Declare.
21698 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
21699 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
21700 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
21701 (aarch64_sve_float_mul_immediate_p): Likewise.
21702 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
21703 rather than an rtx.
21704 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
21705 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
21706 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
21707 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
21708 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
21709 (aarch64_regmode_natural_size): Likewise.
21710 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
21711 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
21713 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
21714 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
21715 for VG and the SVE predicate registers.
21716 (V_ALIASES): Add a "z"-prefixed alias.
21717 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
21718 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
21719 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
21720 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
21721 (REG_CLASS_NAMES): Add entries for them.
21722 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
21723 and the predicate registers.
21724 (aarch64_sve_vg): Declare.
21725 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
21726 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
21727 (REGMODE_NATURAL_SIZE): Define.
21728 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
21730 * config/aarch64/aarch64.c: Include cfgrtl.h.
21731 (simd_immediate_info): Add a constructor for series vectors,
21732 and an associated step field.
21733 (aarch64_sve_vg): New variable.
21734 (aarch64_dbx_register_number): Handle VG and the predicate registers.
21735 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
21736 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
21737 (VEC_ANY_DATA, VEC_STRUCT): New constants.
21738 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
21739 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
21740 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
21741 (aarch64_get_mask_mode): New functions.
21742 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
21743 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
21744 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
21745 predicate modes and predicate registers. Explicitly restrict
21746 GPRs to modes of 16 bytes or smaller. Only allow FP registers
21747 to store a vector mode if it is recognized by
21748 aarch64_classify_vector_mode.
21749 (aarch64_regmode_natural_size): New function.
21750 (aarch64_hard_regno_caller_save_mode): Return the original mode
21752 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
21753 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
21754 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
21755 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
21757 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
21758 does not overlap dest if the function is frame-related. Handle
21760 (aarch64_split_add_offset): New function.
21761 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
21762 them aarch64_add_offset.
21763 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
21764 and update call to aarch64_sub_sp.
21765 (aarch64_add_cfa_expression): New function.
21766 (aarch64_expand_prologue): Pass extra temporary registers to the
21767 functions above. Handle the case in which we need to emit new
21768 DW_CFA_expressions for registers that were originally saved
21769 relative to the stack pointer, but now have to be expressed
21770 relative to the frame pointer.
21771 (aarch64_output_mi_thunk): Pass extra temporary registers to the
21773 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
21774 IP0 and IP1 values for SVE frames.
21775 (aarch64_expand_vec_series): New function.
21776 (aarch64_expand_sve_widened_duplicate): Likewise.
21777 (aarch64_expand_sve_const_vector): Likewise.
21778 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
21779 Handle SVE constants. Use emit_move_insn to move a force_const_mem
21780 into the register, rather than emitting a SET directly.
21781 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
21782 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
21783 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
21784 (offset_9bit_signed_scaled_p): New functions.
21785 (aarch64_replicate_bitmask_imm): New function.
21786 (aarch64_bitmask_imm): Use it.
21787 (aarch64_cannot_force_const_mem): Reject expressions involving
21788 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
21789 (aarch64_classify_index): Handle SVE indices, by requiring
21790 a plain register index with a scale that matches the element size.
21791 (aarch64_classify_address): Handle SVE addresses. Assert that
21792 the mode of the address is VOIDmode or an integer mode.
21793 Update call to aarch64_classify_symbol.
21794 (aarch64_classify_symbolic_expression): Update call to
21795 aarch64_classify_symbol.
21796 (aarch64_const_vec_all_in_range_p): New function.
21797 (aarch64_print_vector_float_operand): Likewise.
21798 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
21799 "vN" for FP registers with SVE modes. Handle (const ...) vectors
21800 and the FP immediates 1.0 and 0.5.
21801 (aarch64_print_address_internal): Handle SVE addresses.
21802 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
21803 (aarch64_regno_regclass): Handle predicate registers.
21804 (aarch64_secondary_reload): Handle big-endian reloads of SVE
21806 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
21807 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
21808 (aarch64_convert_sve_vector_bits): New function.
21809 (aarch64_override_options): Use it to handle -msve-vector-bits=.
21810 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
21811 rather than an rtx.
21812 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
21813 Handle SVE vector and predicate modes. Accept VL-based constants
21814 that need only one temporary register, and VL offsets that require
21815 no temporary registers.
21816 (aarch64_conditional_register_usage): Mark the predicate registers
21817 as fixed if SVE isn't available.
21818 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
21819 Return true for SVE vector and predicate modes.
21820 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
21821 rather than an unsigned int. Handle SVE modes.
21822 (aarch64_preferred_simd_mode): Update call accordingly. Handle
21824 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
21826 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
21827 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
21828 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
21829 (aarch64_sve_float_mul_immediate_p): New functions.
21830 (aarch64_sve_valid_immediate): New function.
21831 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
21832 Explicitly reject structure modes. Check for INDEX constants.
21833 Handle PTRUE and PFALSE constants.
21834 (aarch64_check_zero_based_sve_index_immediate): New function.
21835 (aarch64_simd_imm_zero_p): Delete.
21836 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
21837 vector modes. Accept constants in the range of CNT[BHWD].
21838 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
21839 ask for an Advanced SIMD mode.
21840 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
21841 (aarch64_simd_vector_alignment): Handle SVE predicates.
21842 (aarch64_vectorize_preferred_vector_alignment): New function.
21843 (aarch64_simd_vector_alignment_reachable): Use it instead of
21845 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
21846 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
21848 (MAX_VECT_LEN): Delete.
21849 (expand_vec_perm_d): Add a vec_flags field.
21850 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
21851 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
21852 (aarch64_evpc_ext): Don't apply a big-endian lane correction
21854 (aarch64_evpc_rev): Rename to...
21855 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
21856 (aarch64_evpc_rev_global): New function.
21857 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
21858 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
21860 (aarch64_evpc_sve_tbl): New function.
21861 (aarch64_expand_vec_perm_const_1): Update after rename of
21862 aarch64_evpc_rev. Handle SVE permutes too, trying
21863 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
21864 than aarch64_evpc_tbl.
21865 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
21866 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
21867 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
21868 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
21869 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
21870 (aarch64_expand_sve_vcond): New functions.
21871 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
21872 of aarch64_vector_mode_p.
21873 (aarch64_dwarf_poly_indeterminate_value): New function.
21874 (aarch64_compute_pressure_classes): Likewise.
21875 (aarch64_can_change_mode_class): Likewise.
21876 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
21877 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
21878 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
21879 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
21880 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
21881 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
21882 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
21883 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
21885 (Dn, Dl, Dr): Accept const as well as const_vector.
21886 (Dz): Likewise. Compare against CONST0_RTX.
21887 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
21888 of "vector" where appropriate.
21889 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
21890 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
21891 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
21892 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
21893 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
21894 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
21895 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
21896 (v_int_equiv): Extend to SVE modes.
21897 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
21899 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
21900 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
21901 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
21902 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
21903 (SVE_COND_FP_CMP): New int iterators.
21904 (perm_hilo): Handle the new unpack unspecs.
21905 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
21907 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
21908 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
21909 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
21910 (aarch64_equality_operator, aarch64_constant_vector_operand)
21911 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
21912 (aarch64_sve_nonimmediate_operand): Likewise.
21913 (aarch64_sve_general_operand): Likewise.
21914 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
21915 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
21916 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
21917 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
21918 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
21919 (aarch64_sve_float_arith_immediate): Likewise.
21920 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
21921 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
21922 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
21923 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
21924 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
21925 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
21926 (aarch64_sve_float_arith_operand): Likewise.
21927 (aarch64_sve_float_arith_with_sub_operand): Likewise.
21928 (aarch64_sve_float_mul_operand): Likewise.
21929 (aarch64_sve_vec_perm_operand): Likewise.
21930 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
21931 (aarch64_mov_operand): Accept const_poly_int and const_vector.
21932 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
21933 as well as const_vector.
21934 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
21935 in file. Use CONST0_RTX and CONSTM1_RTX.
21936 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
21937 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
21938 Use aarch64_simd_imm_zero.
21939 * config/aarch64/aarch64-sve.md: New file.
21940 * config/aarch64/aarch64.md: Include it.
21941 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
21942 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
21943 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
21944 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
21945 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
21946 (sve): New attribute.
21947 (enabled): Disable instructions with the sve attribute unless
21949 (movqi, movhi): Pass CONST_POLY_INT operaneds through
21950 aarch64_expand_mov_immediate.
21951 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
21952 CNT[BHSD] immediates.
21953 (movti): Split CONST_POLY_INT moves into two halves.
21954 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
21955 Split additions that need a temporary here if the destination
21956 is the stack pointer.
21957 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
21958 (*add<mode>3_poly_1): New instruction.
21959 (set_clobber_cc): New expander.
21961 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21963 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
21964 parameter and use it instead of GET_MODE_SIZE (innermode). Use
21965 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
21966 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
21967 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
21968 Change innermode from fixed_mode_size to machine_mode.
21969 (simplify_subreg): Update call accordingly. Handle a constant-sized
21970 subreg of a variable-length CONST_VECTOR.
21972 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21973 Alan Hayward <alan.hayward@arm.com>
21974 David Sherwood <david.sherwood@arm.com>
21976 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
21977 (add_offset_to_base): New function, split out from...
21978 (create_mem_ref): ...here. When handling a scale other than 1,
21979 check first whether the address is valid without the offset.
21980 Add it into the base if so, leaving the index and scale as-is.
21982 2018-01-12 Jakub Jelinek <jakub@redhat.com>
21985 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
21986 fold_for_warn before checking if arg2 is INTEGER_CST.
21988 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
21990 * config/rs6000/predicates.md (load_multiple_operation): Delete.
21991 (store_multiple_operation): Delete.
21992 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
21993 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
21994 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
21995 guarded by TARGET_STRING.
21996 (rs6000_output_load_multiple): Delete.
21997 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
21998 OPTION_MASK_STRING / TARGET_STRING handling.
21999 (print_operand) <'N', 'O'>: Add comment that these are unused now.
22000 (const rs6000_opt_masks) <"string">: Change mask to 0.
22001 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
22002 (MASK_STRING): Delete.
22003 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
22005 (load_multiple): Delete.
22012 (store_multiple): Delete.
22019 (movmemsi_8reg): Delete.
22020 (corresponding unnamed define_insn): Delete.
22021 (movmemsi_6reg): Delete.
22022 (corresponding unnamed define_insn): Delete.
22023 (movmemsi_4reg): Delete.
22024 (corresponding unnamed define_insn): Delete.
22025 (movmemsi_2reg): Delete.
22026 (corresponding unnamed define_insn): Delete.
22027 (movmemsi_1reg): Delete.
22028 (corresponding unnamed define_insn): Delete.
22029 * config/rs6000/rs6000.opt (mno-string): New.
22030 (mstring): Replace by deprecation warning stub.
22031 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
22033 2018-01-12 Jakub Jelinek <jakub@redhat.com>
22035 * regrename.c (regrename_do_replace): If replacing the same
22036 reg multiple times, try to reuse last created gen_raw_REG.
22039 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
22040 main to workaround a bug in GDB.
22042 2018-01-12 Tom de Vries <tom@codesourcery.com>
22045 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
22047 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
22049 PR rtl-optimization/80481
22050 * ira-color.c (get_cap_member): New function.
22051 (allocnos_conflict_by_live_ranges_p): Use it.
22052 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
22053 (setup_slot_coalesced_allocno_live_ranges): Ditto.
22055 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
22058 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
22059 (*saddl_se_1): Ditto.
22060 (*ssubsi_1): Ditto.
22061 (*ssubl_se_1): Ditto.
22063 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
22065 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
22066 rather than wi::to_widest for DR_INITs.
22067 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
22068 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
22069 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
22071 (vect_analyze_group_access_1): Note that here.
22073 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
22075 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
22076 polynomial type sizes.
22078 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
22080 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
22081 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
22082 (gimple_add_tmp_var): Likewise.
22084 2018-01-12 Martin Liska <mliska@suse.cz>
22086 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
22087 (gimple_alloc_sizes): Likewise.
22088 (dump_gimple_statistics): Use PRIu64 in printf format.
22089 * gimple.h: Change uint64_t to int.
22091 2018-01-12 Martin Liska <mliska@suse.cz>
22093 * tree-core.h: Use uint64_t instead of int.
22094 * tree.c (tree_node_counts): Likewise.
22095 (tree_node_sizes): Likewise.
22096 (dump_tree_statistics): Use PRIu64 in printf format.
22098 2018-01-12 Martin Liska <mliska@suse.cz>
22100 * Makefile.in: As qsort_chk is implemented in vec.c, add
22101 vec.o to linkage of gencfn-macros.
22102 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
22103 passing the info to record_node_allocation_statistics.
22104 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
22106 * ggc-common.c (struct ggc_usage): Add operator== and use
22107 it in operator< and compare function.
22108 * mem-stats.h (struct mem_usage): Likewise.
22109 * vec.c (struct vec_usage): Remove operator< and compare
22110 function. Can be simply inherited.
22112 2018-01-12 Martin Jambor <mjambor@suse.cz>
22115 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
22116 * tree-ssa-math-opts.c: Include domwalk.h.
22117 (convert_mult_to_fma_1): New function.
22118 (fma_transformation_info): New type.
22119 (fma_deferring_state): Likewise.
22120 (cancel_fma_deferring): New function.
22121 (result_of_phi): Likewise.
22122 (last_fma_candidate_feeds_initial_phi): Likewise.
22123 (convert_mult_to_fma): Added deferring logic, split actual
22124 transformation to convert_mult_to_fma_1.
22125 (math_opts_dom_walker): New type.
22126 (math_opts_dom_walker::after_dom_children): New method, body moved
22127 here from pass_optimize_widening_mul::execute, added deferring logic
22129 (pass_optimize_widening_mul::execute): Moved most of code to
22130 math_opts_dom_walker::after_dom_children.
22131 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
22132 * config/i386/i386.c (ix86_option_override_internal): Added
22133 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
22135 2018-01-12 Richard Biener <rguenther@suse.de>
22138 * dwarf2out.c (gen_variable_die): Do not reset old_die for
22139 inline instance vars.
22141 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
22144 * config/rx/rx.c (rx_is_restricted_memory_address):
22145 Handle SUBREG case.
22147 2018-01-12 Richard Biener <rguenther@suse.de>
22149 PR tree-optimization/80846
22150 * target.def (split_reduction): New target hook.
22151 * targhooks.c (default_split_reduction): New function.
22152 * targhooks.h (default_split_reduction): Declare.
22153 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
22154 target requests first reduce vectors by combining low and high
22156 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
22157 (get_vectype_for_scalar_type_and_size): Export.
22158 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
22159 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
22160 * doc/tm.texi: Regenerate.
22161 * config/i386/i386.c (ix86_split_reduction): Implement
22162 TARGET_VECTORIZE_SPLIT_REDUCTION.
22164 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
22167 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
22168 in PIC mode except for TARGET_VXWORKS_RTP.
22169 * config/sparc/sparc.c: Include cfgrtl.h.
22170 (TARGET_INIT_PIC_REG): Define.
22171 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
22172 (sparc_pic_register_p): New predicate.
22173 (sparc_legitimate_address_p): Use it.
22174 (sparc_legitimize_pic_address): Likewise.
22175 (sparc_delegitimize_address): Likewise.
22176 (sparc_mode_dependent_address_p): Likewise.
22177 (gen_load_pcrel_sym): Remove 4th parameter.
22178 (load_got_register): Adjust call to above. Remove obsolete stuff.
22179 (sparc_expand_prologue): Do not call load_got_register here.
22180 (sparc_flat_expand_prologue): Likewise.
22181 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
22182 (sparc_use_pseudo_pic_reg): New function.
22183 (sparc_init_pic_reg): Likewise.
22184 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
22185 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
22187 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
22189 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
22190 Add item for branch_cost.
22192 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
22194 PR rtl-optimization/83565
22195 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
22196 not extend the result to a larger mode for rotate operations.
22197 (num_sign_bit_copies1): Likewise.
22199 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22202 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
22204 Use values-Xc.o for -pedantic.
22205 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
22207 2018-01-12 Martin Liska <mliska@suse.cz>
22210 * ipa-devirt.c (final_warning_record::grow_type_warnings):
22212 (possible_polymorphic_call_targets): Use it.
22213 (ipa_devirt): Likewise.
22215 2018-01-12 Martin Liska <mliska@suse.cz>
22217 * profile-count.h (enum profile_quality): Use 0 as invalid
22218 enum value of profile_quality.
22220 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
22222 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
22223 -mext-string options.
22225 2018-01-12 Richard Biener <rguenther@suse.de>
22227 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
22228 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
22229 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
22231 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
22233 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
22235 * configure.ac (--with-long-double-format): Add support for the
22236 configuration option to change the default long double format on
22238 * config.gcc (powerpc*-linux*-*): Likewise.
22239 * configure: Regenerate.
22240 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
22241 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
22242 used without modification.
22244 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22246 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
22247 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
22248 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
22249 MISC_BUILTIN_SPEC_BARRIER.
22250 (rs6000_init_builtins): Likewise.
22251 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
22253 (speculation_barrier): New define_insn.
22254 * doc/extend.texi: Document __builtin_speculation_barrier.
22256 2018-01-11 Jakub Jelinek <jakub@redhat.com>
22259 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
22260 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
22261 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
22263 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
22264 integral modes instead of "ss" and "sd".
22265 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
22266 vectors with 32-bit and 64-bit elements.
22267 (vecdupssescalarmodesuffix): New mode attribute.
22268 (vec_dup<mode>): Use it.
22270 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
22273 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
22274 frame if argument is passed on stack.
22276 2018-01-11 Jakub Jelinek <jakub@redhat.com>
22279 * ree.c (combine_reaching_defs): Optimize also
22280 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
22281 reg2=any_extend(exp); reg1=reg2;, formatting fix.
22283 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
22285 PR middle-end/83189
22286 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
22288 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
22290 PR middle-end/83718
22291 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
22292 after they are computed.
22294 2018-01-11 Bin Cheng <bin.cheng@arm.com>
22296 PR tree-optimization/83695
22297 * gimple-loop-linterchange.cc
22298 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
22299 reset cached scev information after interchange.
22300 (pass_linterchange::execute): Remove call to scev_reset_htab.
22302 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22304 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
22305 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
22306 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
22307 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
22308 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
22309 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
22310 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
22311 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
22312 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
22313 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
22314 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
22315 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
22316 (V_lane_reg): Likewise.
22317 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
22319 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
22320 (vfmal_lane_low<mode>_intrinsic,
22321 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
22322 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
22323 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
22324 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
22325 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
22326 vfmsl_lane_high<mode>_intrinsic): New define_insns.
22328 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22330 * config/arm/arm-cpus.in (fp16fml): New feature.
22331 (ALL_SIMD): Add fp16fml.
22332 (armv8.2-a): Add fp16fml as an option.
22333 (armv8.3-a): Likewise.
22334 (armv8.4-a): Add fp16fml as part of fp16.
22335 * config/arm/arm.h (TARGET_FP16FML): Define.
22336 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
22338 * config/arm/arm-modes.def (V2HF): Define.
22339 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
22340 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
22341 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
22342 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
22343 vfmsl_low, vfmsl_high): New set of builtins.
22344 * config/arm/iterators.md (PLUSMINUS): New code iterator.
22345 (vfml_op): New code attribute.
22346 (VFMLHALVES): New int iterator.
22347 (VFML, VFMLSEL): New mode attributes.
22348 (V_reg): Define mapping for V2HF.
22349 (V_hi, V_lo): New mode attributes.
22350 (VF_constraint): Likewise.
22351 (vfml_half, vfml_half_selector): New int attributes.
22352 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
22354 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
22355 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
22357 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
22358 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
22359 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
22360 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
22362 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
22363 Document new effective target and option set.
22365 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22367 * config/arm/arm-cpus.in (armv8_4): New feature.
22368 (ARMv8_4a): New fgroup.
22369 (armv8.4-a): New arch.
22370 * config/arm/arm-tables.opt: Regenerate.
22371 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
22372 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
22373 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
22374 Add matching rules for -march=armv8.4-a and extensions.
22375 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
22377 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
22380 * config/rx/rx.md (BW): New mode attribute.
22381 (sync_lock_test_and_setsi): Add mode suffix to insn output.
22383 2018-01-11 Richard Biener <rguenther@suse.de>
22385 PR tree-optimization/83435
22386 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
22387 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
22388 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
22390 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22391 Alan Hayward <alan.hayward@arm.com>
22392 David Sherwood <david.sherwood@arm.com>
22394 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
22396 (aarch64_classify_address): Initialize it. Track polynomial offsets.
22397 (aarch64_print_address_internal): Use it to check for a zero offset.
22399 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22400 Alan Hayward <alan.hayward@arm.com>
22401 David Sherwood <david.sherwood@arm.com>
22403 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
22404 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
22405 Return a poly_int64 rather than a HOST_WIDE_INT.
22406 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
22407 rather than a HOST_WIDE_INT.
22408 * config/aarch64/aarch64.h (aarch64_frame): Protect with
22409 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
22410 hard_fp_offset, frame_size, initial_adjust, callee_offset and
22411 final_offset from HOST_WIDE_INT to poly_int64.
22412 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
22413 to_constant when getting the number of units in an Advanced SIMD
22415 (aarch64_builtin_vectorized_function): Check for a constant number
22417 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
22419 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
22420 attribute instead of GET_MODE_NUNITS.
22421 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
22422 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
22423 GET_MODE_SIZE for fixed-size registers.
22424 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
22425 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
22426 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
22427 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
22428 (aarch64_print_operand, aarch64_print_address_internal)
22429 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
22430 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
22431 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
22432 Handle polynomial GET_MODE_SIZE.
22433 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
22434 wider than SImode without modification.
22435 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
22436 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
22437 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
22438 passing and returning SVE modes.
22439 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
22440 rather than GEN_INT.
22441 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
22442 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
22443 (aarch64_allocate_and_probe_stack_space): Likewise.
22444 (aarch64_layout_frame): Cope with polynomial offsets.
22445 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
22446 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
22447 polynomial offsets.
22448 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
22449 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
22450 poly_int64 rather than a HOST_WIDE_INT.
22451 (aarch64_get_separate_components, aarch64_process_components)
22452 (aarch64_expand_prologue, aarch64_expand_epilogue)
22453 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
22454 (aarch64_anchor_offset): New function, split out from...
22455 (aarch64_legitimize_address): ...here.
22456 (aarch64_builtin_vectorization_cost): Handle polynomial
22457 TYPE_VECTOR_SUBPARTS.
22458 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
22460 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
22461 number of elements from the PARALLEL rather than the mode.
22462 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
22463 rather than GET_MODE_BITSIZE.
22464 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
22465 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
22466 (aarch64_expand_vec_perm_const_1): Handle polynomial
22467 d->perm.length () and d->perm elements.
22468 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
22469 Apply to_constant to d->perm elements.
22470 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
22471 polynomial CONST_VECTOR_NUNITS.
22472 (aarch64_move_pointer): Take amount as a poly_int64 rather
22474 (aarch64_progress_pointer): Avoid temporary variable.
22475 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
22476 the mode attribute instead of GET_MODE.
22478 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22479 Alan Hayward <alan.hayward@arm.com>
22480 David Sherwood <david.sherwood@arm.com>
22482 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
22483 x exists before using it.
22484 (aarch64_add_constant_internal): Rename to...
22485 (aarch64_add_offset_1): ...this. Replace regnum with separate
22486 src and dest rtxes. Handle the case in which they're different,
22487 including when the offset is zero. Replace scratchreg with an rtx.
22488 Use 2 additions if there is no spare register into which we can
22489 move a 16-bit constant.
22490 (aarch64_add_constant): Delete.
22491 (aarch64_add_offset): Replace reg with separate src and dest
22492 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
22493 Use aarch64_add_offset_1.
22494 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
22495 an rtx rather than an int. Take the delta as a poly_int64
22496 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
22497 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
22498 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
22499 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
22500 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
22501 and aarch64_add_sp.
22502 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
22503 aarch64_add_constant.
22505 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22507 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22508 Use scalar_float_mode.
22510 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22512 * config/aarch64/aarch64-simd.md
22513 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
22514 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
22515 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
22516 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
22517 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
22518 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
22519 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
22520 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
22521 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
22522 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
22524 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22527 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
22528 targ_options->x_arm_arch_string is non NULL.
22530 2018-01-11 Tamar Christina <tamar.christina@arm.com>
22532 * config/aarch64/aarch64.h
22533 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
22535 2018-01-11 Sudakshina Das <sudi.das@arm.com>
22538 * expmed.c (emit_store_flag_force): Swap if const op0
22539 and change VOIDmode to mode of op0.
22541 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22543 PR rtl-optimization/83761
22544 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
22545 than bytes to mode_for_size.
22547 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
22549 PR middle-end/83189
22550 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
22551 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
22554 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
22556 PR middle-end/83575
22557 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
22558 when in layout mode.
22559 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
22560 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
22563 2018-01-10 Michael Collison <michael.collison@arm.com>
22565 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
22566 * config/aarch64/aarch64-option-extension.def: Add
22567 AARCH64_OPT_EXTENSION of 'fp16fml'.
22568 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22569 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
22570 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
22571 * config/aarch64/constraints.md (Ui7): New constraint.
22572 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
22573 (VFMLA_SEL_W): Ditto.
22576 (VFMLA16_LOW): New int iterator.
22577 (VFMLA16_HIGH): Ditto.
22578 (UNSPEC_FMLAL): New unspec.
22579 (UNSPEC_FMLSL): Ditto.
22580 (UNSPEC_FMLAL2): Ditto.
22581 (UNSPEC_FMLSL2): Ditto.
22582 (f16mac): New code attribute.
22583 * config/aarch64/aarch64-simd-builtins.def
22584 (aarch64_fmlal_lowv2sf): Ditto.
22585 (aarch64_fmlsl_lowv2sf): Ditto.
22586 (aarch64_fmlalq_lowv4sf): Ditto.
22587 (aarch64_fmlslq_lowv4sf): Ditto.
22588 (aarch64_fmlal_highv2sf): Ditto.
22589 (aarch64_fmlsl_highv2sf): Ditto.
22590 (aarch64_fmlalq_highv4sf): Ditto.
22591 (aarch64_fmlslq_highv4sf): Ditto.
22592 (aarch64_fmlal_lane_lowv2sf): Ditto.
22593 (aarch64_fmlsl_lane_lowv2sf): Ditto.
22594 (aarch64_fmlal_laneq_lowv2sf): Ditto.
22595 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
22596 (aarch64_fmlalq_lane_lowv4sf): Ditto.
22597 (aarch64_fmlsl_lane_lowv4sf): Ditto.
22598 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
22599 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
22600 (aarch64_fmlal_lane_highv2sf): Ditto.
22601 (aarch64_fmlsl_lane_highv2sf): Ditto.
22602 (aarch64_fmlal_laneq_highv2sf): Ditto.
22603 (aarch64_fmlsl_laneq_highv2sf): Ditto.
22604 (aarch64_fmlalq_lane_highv4sf): Ditto.
22605 (aarch64_fmlsl_lane_highv4sf): Ditto.
22606 (aarch64_fmlalq_laneq_highv4sf): Ditto.
22607 (aarch64_fmlsl_laneq_highv4sf): Ditto.
22608 * config/aarch64/aarch64-simd.md:
22609 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
22610 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
22611 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
22612 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
22613 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
22614 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
22615 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
22616 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
22617 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
22618 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
22619 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
22620 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
22621 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
22622 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
22623 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
22624 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
22625 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
22626 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
22627 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
22628 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
22629 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
22630 (vfmlsl_low_u32): Ditto.
22631 (vfmlalq_low_u32): Ditto.
22632 (vfmlslq_low_u32): Ditto.
22633 (vfmlal_high_u32): Ditto.
22634 (vfmlsl_high_u32): Ditto.
22635 (vfmlalq_high_u32): Ditto.
22636 (vfmlslq_high_u32): Ditto.
22637 (vfmlal_lane_low_u32): Ditto.
22638 (vfmlsl_lane_low_u32): Ditto.
22639 (vfmlal_laneq_low_u32): Ditto.
22640 (vfmlsl_laneq_low_u32): Ditto.
22641 (vfmlalq_lane_low_u32): Ditto.
22642 (vfmlslq_lane_low_u32): Ditto.
22643 (vfmlalq_laneq_low_u32): Ditto.
22644 (vfmlslq_laneq_low_u32): Ditto.
22645 (vfmlal_lane_high_u32): Ditto.
22646 (vfmlsl_lane_high_u32): Ditto.
22647 (vfmlal_laneq_high_u32): Ditto.
22648 (vfmlsl_laneq_high_u32): Ditto.
22649 (vfmlalq_lane_high_u32): Ditto.
22650 (vfmlslq_lane_high_u32): Ditto.
22651 (vfmlalq_laneq_high_u32): Ditto.
22652 (vfmlslq_laneq_high_u32): Ditto.
22653 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
22654 (AARCH64_FL_FOR_ARCH8_4): New.
22655 (AARCH64_ISA_F16FML): New ISA flag.
22656 (TARGET_F16FML): New feature flag for fp16fml.
22657 (doc/invoke.texi): Document new fp16fml option.
22659 2018-01-10 Michael Collison <michael.collison@arm.com>
22661 * config/aarch64/aarch64-builtins.c:
22662 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
22663 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22664 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
22665 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
22666 (AARCH64_ISA_SHA3): New ISA flag.
22667 (TARGET_SHA3): New feature flag for sha3.
22668 * config/aarch64/iterators.md (sha512_op): New int attribute.
22669 (CRYPTO_SHA512): New int iterator.
22670 (UNSPEC_SHA512H): New unspec.
22671 (UNSPEC_SHA512H2): Ditto.
22672 (UNSPEC_SHA512SU0): Ditto.
22673 (UNSPEC_SHA512SU1): Ditto.
22674 * config/aarch64/aarch64-simd-builtins.def
22675 (aarch64_crypto_sha512hqv2di): New builtin.
22676 (aarch64_crypto_sha512h2qv2di): Ditto.
22677 (aarch64_crypto_sha512su0qv2di): Ditto.
22678 (aarch64_crypto_sha512su1qv2di): Ditto.
22679 (aarch64_eor3qv8hi): Ditto.
22680 (aarch64_rax1qv2di): Ditto.
22681 (aarch64_xarqv2di): Ditto.
22682 (aarch64_bcaxqv8hi): Ditto.
22683 * config/aarch64/aarch64-simd.md:
22684 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
22685 (aarch64_crypto_sha512su0qv2di): Ditto.
22686 (aarch64_crypto_sha512su1qv2di): Ditto.
22687 (aarch64_eor3qv8hi): Ditto.
22688 (aarch64_rax1qv2di): Ditto.
22689 (aarch64_xarqv2di): Ditto.
22690 (aarch64_bcaxqv8hi): Ditto.
22691 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
22692 (vsha512h2q_u64): Ditto.
22693 (vsha512su0q_u64): Ditto.
22694 (vsha512su1q_u64): Ditto.
22695 (veor3q_u16): Ditto.
22696 (vrax1q_u64): Ditto.
22697 (vxarq_u64): Ditto.
22698 (vbcaxq_u16): Ditto.
22699 * config/arm/types.md (crypto_sha512): New type attribute.
22700 (crypto_sha3): Ditto.
22701 (doc/invoke.texi): Document new sha3 option.
22703 2018-01-10 Michael Collison <michael.collison@arm.com>
22705 * config/aarch64/aarch64-builtins.c:
22706 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
22707 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22708 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
22709 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
22710 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
22711 (AARCH64_ISA_SM4): New ISA flag.
22712 (TARGET_SM4): New feature flag for sm4.
22713 * config/aarch64/aarch64-simd-builtins.def
22714 (aarch64_sm3ss1qv4si): Ditto.
22715 (aarch64_sm3tt1aq4si): Ditto.
22716 (aarch64_sm3tt1bq4si): Ditto.
22717 (aarch64_sm3tt2aq4si): Ditto.
22718 (aarch64_sm3tt2bq4si): Ditto.
22719 (aarch64_sm3partw1qv4si): Ditto.
22720 (aarch64_sm3partw2qv4si): Ditto.
22721 (aarch64_sm4eqv4si): Ditto.
22722 (aarch64_sm4ekeyqv4si): Ditto.
22723 * config/aarch64/aarch64-simd.md:
22724 (aarch64_sm3ss1qv4si): Ditto.
22725 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
22726 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
22727 (aarch64_sm4eqv4si): Ditto.
22728 (aarch64_sm4ekeyqv4si): Ditto.
22729 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
22730 (sm3part_op): Ditto.
22731 (CRYPTO_SM3TT): Ditto.
22732 (CRYPTO_SM3PART): Ditto.
22733 (UNSPEC_SM3SS1): New unspec.
22734 (UNSPEC_SM3TT1A): Ditto.
22735 (UNSPEC_SM3TT1B): Ditto.
22736 (UNSPEC_SM3TT2A): Ditto.
22737 (UNSPEC_SM3TT2B): Ditto.
22738 (UNSPEC_SM3PARTW1): Ditto.
22739 (UNSPEC_SM3PARTW2): Ditto.
22740 (UNSPEC_SM4E): Ditto.
22741 (UNSPEC_SM4EKEY): Ditto.
22742 * config/aarch64/constraints.md (Ui2): New constraint.
22743 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
22744 * config/arm/types.md (crypto_sm3): New type attribute.
22745 (crypto_sm4): Ditto.
22746 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
22747 (vsm3tt1aq_u32): Ditto.
22748 (vsm3tt1bq_u32): Ditto.
22749 (vsm3tt2aq_u32): Ditto.
22750 (vsm3tt2bq_u32): Ditto.
22751 (vsm3partw1q_u32): Ditto.
22752 (vsm3partw2q_u32): Ditto.
22753 (vsm4eq_u32): Ditto.
22754 (vsm4ekeyq_u32): Ditto.
22755 (doc/invoke.texi): Document new sm4 option.
22757 2018-01-10 Michael Collison <michael.collison@arm.com>
22759 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
22760 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
22761 (AARCH64_FL_FOR_ARCH8_4): New.
22762 (AARCH64_FL_V8_4): New flag.
22763 (doc/invoke.texi): Document new armv8.4-a option.
22765 2018-01-10 Michael Collison <michael.collison@arm.com>
22767 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22768 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
22769 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
22770 * config/aarch64/aarch64-option-extension.def: Add
22771 AARCH64_OPT_EXTENSION of 'sha2'.
22772 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
22773 (crypto): Disable sha2 and aes if crypto disabled.
22774 (crypto): Enable aes and sha2 if enabled.
22775 (simd): Disable sha2 and aes if simd disabled.
22776 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
22778 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
22779 (TARGET_SHA2): New feature flag for sha2.
22780 (TARGET_AES): New feature flag for aes.
22781 * config/aarch64/aarch64-simd.md:
22782 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
22783 conditional on TARGET_AES.
22784 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
22785 (aarch64_crypto_sha1hsi): Make pattern conditional
22787 (aarch64_crypto_sha1hv4si): Ditto.
22788 (aarch64_be_crypto_sha1hv4si): Ditto.
22789 (aarch64_crypto_sha1su1v4si): Ditto.
22790 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
22791 (aarch64_crypto_sha1su0v4si): Ditto.
22792 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
22793 (aarch64_crypto_sha256su0v4si): Ditto.
22794 (aarch64_crypto_sha256su1v4si): Ditto.
22795 (doc/invoke.texi): Document new aes and sha2 options.
22797 2018-01-10 Martin Sebor <msebor@redhat.com>
22799 PR tree-optimization/83781
22800 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
22803 2018-01-11 Martin Sebor <msebor@gmail.com>
22804 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22806 PR tree-optimization/83501
22807 PR tree-optimization/81703
22809 * tree-ssa-strlen.c (get_string_cst): Rename...
22810 (get_string_len): ...to this. Handle global constants.
22811 (handle_char_store): Adjust.
22813 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
22814 Jim Wilson <jimw@sifive.com>
22816 * config/riscv/riscv-protos.h (riscv_output_return): New.
22817 * config/riscv/riscv.c (struct machine_function): New naked_p field.
22818 (riscv_attribute_table, riscv_output_return),
22819 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
22820 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
22821 (riscv_compute_frame_info): Only compute frame->mask if not a naked
22823 (riscv_expand_prologue): Add early return for naked function.
22824 (riscv_expand_epilogue): Likewise.
22825 (riscv_function_ok_for_sibcall): Return false for naked function.
22826 (riscv_set_current_function): New.
22827 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
22828 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
22829 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
22830 * doc/extend.texi (RISC-V Function Attributes): New.
22832 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
22834 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
22835 check for 128-bit long double before checking TCmode.
22836 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
22837 128-bit long doubles before checking TFmode or TCmode.
22838 (FLOAT128_IBM_P): Likewise.
22840 2018-01-10 Martin Sebor <msebor@redhat.com>
22842 PR tree-optimization/83671
22843 * builtins.c (c_strlen): Unconditionally return zero for the empty
22845 Use -Warray-bounds for warnings.
22846 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
22847 for non-constant array indices with COMPONENT_REF, arrays of
22848 arrays, and pointers to arrays.
22849 (gimple_fold_builtin_strlen): Determine and set length range for
22850 non-constant character arrays.
22852 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
22854 PR middle-end/81897
22855 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
22858 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
22860 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
22862 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
22865 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
22866 VECTOR_MEM_ALTIVEC_OR_VSX_P.
22867 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
22868 indexed_or_indirect_operand predicate.
22869 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
22870 (*vsx_le_perm_load_v8hi): Likewise.
22871 (*vsx_le_perm_load_v16qi): Likewise.
22872 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
22873 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
22874 (*vsx_le_perm_store_v8hi): Likewise.
22875 (*vsx_le_perm_store_v16qi): Likewise.
22876 (eight unnamed splitters): Likewise.
22878 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
22880 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
22881 * config/rs6000/emmintrin.h: Likewise.
22882 * config/rs6000/mmintrin.h: Likewise.
22883 * config/rs6000/xmmintrin.h: Likewise.
22885 2018-01-10 David Malcolm <dmalcolm@redhat.com>
22888 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
22890 * tree.c (tree_nop_conversion): Return true for location wrapper
22892 (maybe_wrap_with_location): New function.
22893 (selftest::check_strip_nops): New function.
22894 (selftest::test_location_wrappers): New function.
22895 (selftest::tree_c_tests): Call it.
22896 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
22897 (maybe_wrap_with_location): New decl.
22898 (EXPR_LOCATION_WRAPPER_P): New macro.
22899 (location_wrapper_p): New inline function.
22900 (tree_strip_any_location_wrapper): New inline function.
22902 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
22905 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
22906 stack_realign_offset for the largest alignment of stack slot
22908 (ix86_find_max_used_stack_alignment): New function.
22909 (ix86_finalize_stack_frame_flags): Use it. Set
22910 max_used_stack_alignment if we don't realign stack.
22911 * config/i386/i386.h (machine_function): Add
22912 max_used_stack_alignment.
22914 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
22916 * config/arm/arm.opt (-mbranch-cost): New option.
22917 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
22920 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
22923 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
22924 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
22926 2018-01-10 Richard Biener <rguenther@suse.de>
22929 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
22930 early out so it also covers the case where we have a non-NULL
22933 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
22935 PR tree-optimization/83753
22936 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
22937 for non-strided grouped accesses if the number of elements is 1.
22939 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
22942 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
22943 * i386.h (TARGET_USE_GATHER): Define.
22944 * x86-tune.def (X86_TUNE_USE_GATHER): New.
22946 2018-01-10 Martin Liska <mliska@suse.cz>
22949 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
22950 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
22952 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
22953 CLEANUP_NO_PARTITIONING is not set.
22955 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
22957 * doc/rtl.texi: Remove documentation of (const ...) wrappers
22958 for vectors, as a partial revert of r254296.
22959 * rtl.h (const_vec_p): Delete.
22960 (const_vec_duplicate_p): Don't test for vector CONSTs.
22961 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
22962 * expmed.c (make_tree): Likewise.
22965 * common.md (E, F): Use CONSTANT_P instead of checking for
22967 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
22968 checking for CONST_VECTOR.
22970 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
22972 PR middle-end/83575
22973 * predict.c (force_edge_cold): Handle in more sane way edges
22974 with no prediction.
22976 2018-01-09 Carl Love <cel@us.ibm.com>
22978 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
22980 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
22981 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
22982 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
22983 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
22984 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
22985 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
22986 * config/rs6000/rs6000-protos.h: Add extern defition for
22987 rs6000_generate_float2_double_code.
22988 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
22990 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
22991 (float2_v2df): Add define_expand.
22993 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
22996 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
22997 op_mode in the force_to_mode call.
22999 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
23001 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
23002 instead of checking each element individually.
23003 (aarch64_evpc_uzp): Likewise.
23004 (aarch64_evpc_zip): Likewise.
23005 (aarch64_evpc_ext): Likewise.
23006 (aarch64_evpc_rev): Likewise.
23007 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
23008 instead of checking each element individually. Return true without
23010 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
23011 whether all selected elements come from the same input, instead of
23012 checking each element individually. Remove calls to gen_rtx_REG,
23013 start_sequence and end_sequence and instead assert that no rtl is
23016 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
23018 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
23019 order of HIGH and CONST checks.
23021 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
23023 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
23024 if the destination isn't an SSA_NAME.
23026 2018-01-09 Richard Biener <rguenther@suse.de>
23028 PR tree-optimization/83668
23029 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
23031 (canonicalize_loop_form): ... here, renamed from ...
23032 (canonicalize_loop_closed_ssa_form): ... this and amended to
23033 swap successor edges for loop exit blocks to make us use
23034 the RPO order we need for initial schedule generation.
23036 2018-01-09 Joseph Myers <joseph@codesourcery.com>
23038 PR tree-optimization/64811
23039 * match.pd: When optimizing comparisons with Inf, avoid
23040 introducing or losing exceptions from comparisons with NaN.
23042 2018-01-09 Martin Liska <mliska@suse.cz>
23045 * asan.c (shadow_mem_size): Add gcc_assert.
23047 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
23049 Don't save registers in main().
23052 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
23053 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
23054 * config/avr/avr.c (avr_set_current_function): Don't error if
23055 naked, OS_task or OS_main are specified at the same time.
23056 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
23058 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
23060 * common/config/avr/avr-common.c (avr_option_optimization_table):
23061 Switch on -mmain-is-OS_task for optimizing compilations.
23063 2018-01-09 Richard Biener <rguenther@suse.de>
23065 PR tree-optimization/83572
23066 * graphite.c: Include cfganal.h.
23067 (graphite_transform_loops): Connect infinite loops to exit
23068 and remove fake edges at the end.
23070 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
23072 * ipa-inline.c (edge_badness): Revert accidental checkin.
23074 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
23077 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
23078 symbols; not inline clones.
23080 2018-01-09 Jakub Jelinek <jakub@redhat.com>
23083 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
23084 hard registers. Formatting fixes.
23086 PR preprocessor/83722
23087 * gcc.c (try_generate_repro): Pass
23088 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
23089 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
23092 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
23093 Kito Cheng <kito.cheng@gmail.com>
23095 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
23096 (riscv_leaf_function_p): Delete.
23097 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
23099 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23101 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
23103 (do_ifelse): New function.
23104 (do_isel): New function.
23105 (do_sub3): New function.
23106 (do_add3): New function.
23107 (do_load_mask_compare): New function.
23108 (do_overlap_load_compare): New function.
23109 (expand_compare_loop): New function.
23110 (expand_block_compare): Call expand_compare_loop() when appropriate.
23111 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
23112 option description.
23113 (-mblock-compare-inline-loop-limit): New option.
23115 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23118 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
23119 Reverse order of second and third operands in first alternative.
23120 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
23121 of first and second elements in UNSPEC_VPERMR vector.
23122 (altivec_expand_vec_perm_le): Likewise.
23124 2018-01-08 Jeff Law <law@redhat.com>
23126 PR rtl-optimizatin/81308
23127 * tree-switch-conversion.c (cfg_altered): New file scoped static.
23128 (process_switch): If group_case_labels makes a change, then set
23130 (pass_convert_switch::execute): If a switch is converted, then
23131 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
23133 PR rtl-optimization/81308
23134 * recog.c (split_all_insns): Conditionally cleanup the CFG after
23137 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
23139 PR target/83663 - Revert r255946
23140 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
23141 generation for cases where splatting a value is not useful.
23142 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
23143 across a vec_duplicate and a paradoxical subreg forming a vector
23144 mode to a vec_concat.
23146 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23148 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
23149 -march=armv8.3-a variants.
23150 * config/arm/t-multilib: Likewise.
23151 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
23153 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23155 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
23157 (cceq_ior_compare_complement): Give it a name so I can use it, and
23158 change boolean_or_operator predicate to boolean_operator so it can
23159 be used to generate a crand.
23160 (eqne): New code iterator.
23161 (bd/bd_neg): New code_attrs.
23162 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
23163 a single define_insn.
23164 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
23165 decrement (bdnzt/bdnzf/bdzt/bdzf).
23166 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
23167 with the new names of the branch decrement patterns, and added the
23168 names of the branch decrement conditional patterns.
23170 2018-01-08 Richard Biener <rguenther@suse.de>
23172 PR tree-optimization/83563
23173 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
23176 2018-01-08 Richard Biener <rguenther@suse.de>
23178 PR middle-end/83713
23179 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
23181 2018-01-08 Richard Biener <rguenther@suse.de>
23183 PR tree-optimization/83685
23184 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
23185 references to abnormals.
23187 2018-01-08 Richard Biener <rguenther@suse.de>
23190 * dwarf2out.c (output_indirect_strings): Handle empty
23191 skeleton_debug_str_hash.
23192 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
23194 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
23196 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
23197 (emit_store_direct): Likewise.
23198 (arc_trampoline_adjust_address): Likewise.
23199 (arc_asm_trampoline_template): New function.
23200 (arc_initialize_trampoline): Use asm_trampoline_template.
23201 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
23202 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
23203 * config/arc/arc.md (flush_icache): Delete pattern.
23205 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
23207 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
23208 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
23211 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
23214 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
23215 by not USED_FOR_TARGET.
23216 (make_pass_resolve_sw_modes): Likewise.
23218 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
23220 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
23223 2018-01-08 Richard Biener <rguenther@suse.de>
23225 PR middle-end/83580
23226 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
23228 2018-01-08 Richard Biener <rguenther@suse.de>
23230 PR middle-end/83517
23231 * match.pd ((t * 2) / 2) -> t): Add missing :c.
23233 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
23235 PR middle-end/81897
23236 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
23237 basic blocks with a small number of successors.
23238 (convert_control_dep_chain_into_preds): Improve handling of
23240 (dump_predicates): Split apart into...
23241 (dump_pred_chain): ...here...
23242 (dump_pred_info): ...and here.
23243 (can_one_predicate_be_invalidated_p): Add debugging printfs.
23244 (can_chain_union_be_invalidated_p): Improve check for invalidation
23246 (uninit_uses_cannot_happen): Avoid unnecessary if
23247 convert_control_dep_chain_into_preds yielded nothing.
23249 2018-01-06 Martin Sebor <msebor@redhat.com>
23251 PR tree-optimization/83640
23252 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
23253 subtracting negative offset from size.
23254 (builtin_access::overlap): Adjust offset bounds of the access to fall
23255 within the size of the object if possible.
23257 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
23259 PR rtl-optimization/83699
23260 * expmed.c (extract_bit_field_1): Restrict the vector usage of
23261 extract_bit_field_as_subreg to cases in which the extracted
23262 value is also a vector.
23264 * lra-constraints.c (process_alt_operands): Test for the equivalence
23265 substitutions when detecting a possible reload cycle.
23267 2018-01-06 Jakub Jelinek <jakub@redhat.com>
23270 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
23271 by default if flag_selective_schedling{,2}. Formatting fixes.
23273 PR rtl-optimization/83682
23274 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
23275 if it has non-VECTOR_MODE element mode.
23276 (vec_duplicate_p): Likewise.
23278 PR middle-end/83694
23279 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
23280 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
23282 2018-01-05 Jakub Jelinek <jakub@redhat.com>
23285 * config/i386/i386-builtin.def
23286 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
23287 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
23288 Require also OPTION_MASK_ISA_AVX512F in addition to
23289 OPTION_MASK_ISA_GFNI.
23290 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
23291 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
23292 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
23293 to OPTION_MASK_ISA_GFNI.
23294 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
23295 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
23296 OPTION_MASK_ISA_AVX512BW.
23297 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
23298 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
23299 addition to OPTION_MASK_ISA_GFNI.
23300 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
23301 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
23302 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
23303 to OPTION_MASK_ISA_GFNI.
23304 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
23305 a requirement for all ISAs rather than any of them with a few
23307 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
23309 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
23310 bitmasks to be enabled with 3 exceptions, instead of requiring any
23311 enabled ISA with lots of exceptions.
23312 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
23313 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
23314 Change avx512bw in isa attribute to avx512f.
23315 * config/i386/sgxintrin.h: Add license boilerplate.
23316 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
23317 to __AVX512F__ and __AVX512VL to __AVX512VL__.
23318 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
23319 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
23321 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
23322 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
23323 temporarily sse2 rather than sse if not enabled already.
23326 * config/i386/sse.md (VI248_VLBW): Rename to ...
23327 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
23328 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
23329 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
23330 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
23331 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
23332 mode iterator instead of VI248_VLBW.
23334 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
23336 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
23337 (record_modified): Skip clobbers; add debug output.
23338 (param_change_prob): Use sreal frequencies.
23340 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
23342 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
23343 punt for user-aligned variables.
23345 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
23347 * tree-chrec.c (chrec_contains_symbols): Return true for
23350 2018-01-05 Sudakshina Das <sudi.das@arm.com>
23353 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
23354 of (x|y) == x for BICS pattern.
23356 2018-01-05 Jakub Jelinek <jakub@redhat.com>
23358 PR tree-optimization/83605
23359 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
23360 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
23363 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
23365 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
23366 * config/epiphany/rtems.h: New file.
23368 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23369 Uros Bizjak <ubizjak@gmail.com>
23372 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
23373 QIreg_operand instead of register_operand predicate.
23374 * config/i386/i386.c (ix86_rop_should_change_byte_p,
23375 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
23376 comments instead of -fmitigate[-_]rop.
23378 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23381 * cgraphunit.c (symbol_table::compile): Switch to text_section
23382 before calling assembly_start debug hook.
23383 * run-rtl-passes.c (run_rtl_passes): Likewise.
23386 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23388 * tree-vrp.c (extract_range_from_binary_expr_1): Check
23389 range_int_cst_p rather than !symbolic_range_p before calling
23390 extract_range_from_multiplicative_op_1.
23392 2018-01-04 Jeff Law <law@redhat.com>
23394 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
23395 redundant test in assertion.
23397 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23399 * doc/rtl.texi: Document machine_mode wrapper classes.
23401 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23403 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
23404 using tree_to_uhwi.
23406 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23408 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
23409 the VEC_PERM_EXPR fold to fail.
23411 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23414 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
23415 to switched_sections.
23417 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23420 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
23421 test for d.testing.
23423 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
23426 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
23427 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
23429 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23432 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
23433 is BLKmode and bitpos not zero or mode change is needed.
23435 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23438 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
23441 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
23444 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
23445 instead of MULT rtx. Update all corresponding splitters.
23446 (*saddl_se): Ditto.
23447 (*ssub<modesuffix>): Ditto.
23448 (*ssubl_se): Ditto.
23449 (*cmp_sadd_di): Update split patterns.
23450 (*cmp_sadd_si): Ditto.
23451 (*cmp_sadd_sidi): Ditto.
23452 (*cmp_ssub_di): Ditto.
23453 (*cmp_ssub_si): Ditto.
23454 (*cmp_ssub_sidi): Ditto.
23455 * config/alpha/predicates.md (const23_operand): New predicate.
23456 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
23457 Look for ASHIFT, not MULT inner operand.
23458 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
23460 2018-01-04 Martin Liska <mliska@suse.cz>
23462 PR gcov-profile/83669
23463 * gcov.c (output_intermediate_file): Add version to intermediate
23465 * doc/gcov.texi: Document new field 'version' in intermediate
23466 file format. Fix location of '-k' option of gcov command.
23468 2018-01-04 Martin Liska <mliska@suse.cz>
23471 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
23473 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23475 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
23477 2018-01-03 Martin Sebor <msebor@redhat.com>
23479 PR tree-optimization/83655
23480 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
23481 checking calls with invalid arguments.
23483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23485 * tree-vect-stmts.c (vect_get_store_rhs): New function.
23486 (vectorizable_mask_load_store): Delete.
23487 (vectorizable_call): Return false for masked loads and stores.
23488 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
23489 instead of gimple_assign_rhs1.
23490 (vectorizable_load): Handle IFN_MASK_LOAD.
23491 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
23493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23495 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
23497 (vectorizable_mask_load_store): ...here.
23498 (vectorizable_load): ...and here.
23500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23502 * tree-vect-stmts.c (vect_build_all_ones_mask)
23503 (vect_build_zero_merge_argument): New functions, split out from...
23504 (vectorizable_load): ...here.
23506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23508 * tree-vect-stmts.c (vect_check_store_rhs): New function,
23510 (vectorizable_mask_load_store): ...here.
23511 (vectorizable_store): ...and here.
23513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23515 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
23517 (vectorizable_mask_load_store): ...here.
23519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23521 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
23522 (vect_model_store_cost): Take a vec_load_store_type instead of a
23524 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
23525 (vect_model_store_cost): Take a vec_load_store_type instead of a
23527 (vectorizable_mask_load_store): Update accordingly.
23528 (vectorizable_store): Likewise.
23529 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
23531 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23533 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
23534 IFN_MASK_LOAD calls here rather than...
23535 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
23537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23538 Alan Hayward <alan.hayward@arm.com>
23539 David Sherwood <david.sherwood@arm.com>
23541 * expmed.c (extract_bit_field_1): For vector extracts,
23542 fall back to extract_bit_field_as_subreg if vec_extract
23545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23546 Alan Hayward <alan.hayward@arm.com>
23547 David Sherwood <david.sherwood@arm.com>
23549 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
23550 they are variable or constant sized.
23551 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
23552 slots for constant-sized data.
23554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23555 Alan Hayward <alan.hayward@arm.com>
23556 David Sherwood <david.sherwood@arm.com>
23558 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
23559 handling COND_EXPRs with boolean comparisons, try to find a better
23560 basis for the mask type than the boolean itself.
23562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23564 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
23565 is calculated and how it can be overridden.
23566 * genmodes.c (max_bitsize_mode_any_mode): New variable.
23567 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
23569 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
23572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23573 Alan Hayward <alan.hayward@arm.com>
23574 David Sherwood <david.sherwood@arm.com>
23576 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
23577 Remove the mode argument.
23578 (aarch64_simd_valid_immediate): Remove the mode and inverse
23580 * config/aarch64/iterators.md (bitsize): New iterator.
23581 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
23582 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
23583 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
23584 aarch64_simd_valid_immediate.
23585 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
23586 (aarch64_reg_or_bic_imm): Likewise.
23587 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
23588 with an insn_type enum and msl with a modifier_type enum.
23589 Replace element_width with a scalar_mode. Change the shift
23590 to unsigned int. Add constructors for scalar_float_mode and
23591 scalar_int_mode elements.
23592 (aarch64_vect_float_const_representable_p): Delete.
23593 (aarch64_can_const_movi_rtx_p)
23594 (aarch64_simd_scalar_immediate_valid_for_move)
23595 (aarch64_simd_make_constant): Update call to
23596 aarch64_simd_valid_immediate.
23597 (aarch64_advsimd_valid_immediate_hs): New function.
23598 (aarch64_advsimd_valid_immediate): Likewise.
23599 (aarch64_simd_valid_immediate): Remove mode and inverse
23600 arguments. Rewrite to use the above. Use const_vec_duplicate_p
23601 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
23602 and aarch64_float_const_representable_p on the result.
23603 (aarch64_output_simd_mov_immediate): Remove mode argument.
23604 Update call to aarch64_simd_valid_immediate and use of
23605 simd_immediate_info.
23606 (aarch64_output_scalar_simd_mov_immediate): Update call
23609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23610 Alan Hayward <alan.hayward@arm.com>
23611 David Sherwood <david.sherwood@arm.com>
23613 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
23614 (mode_nunits): Likewise CONST_MODE_NUNITS.
23615 * machmode.def (ADJUST_NUNITS): Document.
23616 * genmodes.c (mode_data::need_nunits_adj): New field.
23617 (blank_mode): Update accordingly.
23618 (adj_nunits): New variable.
23619 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
23621 (emit_mode_size_inline): Set need_bytesize_adj for all modes
23622 listed in adj_nunits.
23623 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
23624 listed in adj_nunits. Don't emit case statements for such modes.
23625 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
23626 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
23627 nothing if adj_nunits is nonnull.
23628 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
23629 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
23630 (emit_mode_fbit): Update use of print_maybe_const_decl.
23631 (emit_move_size): Likewise. Treat the array as non-const
23633 (emit_mode_adjustments): Handle adj_nunits.
23635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23637 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
23638 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
23639 (VECTOR_MODES): Use it.
23640 (make_vector_modes): Take the prefix as an argument.
23642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23643 Alan Hayward <alan.hayward@arm.com>
23644 David Sherwood <david.sherwood@arm.com>
23646 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
23647 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
23648 for MODE_VECTOR_BOOL.
23649 * machmode.def (VECTOR_BOOL_MODE): Document.
23650 * genmodes.c (VECTOR_BOOL_MODE): New macro.
23651 (make_vector_bool_mode): New function.
23652 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
23654 * lto-streamer-in.c (lto_input_mode_table): Likewise.
23655 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
23657 * stor-layout.c (int_mode_for_mode): Likewise.
23658 * tree.c (build_vector_type_for_mode): Likewise.
23659 * varasm.c (output_constant_pool_2): Likewise.
23660 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
23661 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
23662 for MODE_VECTOR_BOOL.
23663 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
23664 of mode class checks.
23665 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
23666 instead of a list of mode class checks.
23667 (expand_vector_scalar_condition): Likewise.
23668 (type_for_widest_vector_mode): Handle BImode as an inner mode.
23670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23671 Alan Hayward <alan.hayward@arm.com>
23672 David Sherwood <david.sherwood@arm.com>
23674 * machmode.h (mode_size): Change from unsigned short to
23676 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
23677 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
23678 or if measurement_type is not polynomial.
23679 (fixed_size_mode::includes_p): Check for constant-sized modes.
23680 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
23681 return a poly_uint16 rather than an unsigned short.
23682 (emit_mode_size): Change the type of mode_size from unsigned short
23683 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
23684 (emit_mode_adjustments): Cope with polynomial vector sizes.
23685 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
23687 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
23689 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
23690 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
23691 * caller-save.c (setup_save_areas): Likewise.
23692 (replace_reg_with_saved_mem): Likewise.
23693 * calls.c (emit_library_call_value_1): Likewise.
23694 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
23695 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
23696 (gen_lowpart_for_combine): Likewise.
23697 * convert.c (convert_to_integer_1): Likewise.
23698 * cse.c (equiv_constant, cse_insn): Likewise.
23699 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
23700 (cselib_subst_to_values): Likewise.
23701 * dce.c (word_dce_process_block): Likewise.
23702 * df-problems.c (df_word_lr_mark_ref): Likewise.
23703 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
23704 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
23705 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
23706 (rtl_for_decl_location): Likewise.
23707 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
23708 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
23709 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
23710 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
23711 (expand_expr_real_1): Likewise.
23712 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
23713 (pad_below): Likewise.
23714 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
23715 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
23716 * ira.c (get_subreg_tracking_sizes): Likewise.
23717 * ira-build.c (ira_create_allocno_objects): Likewise.
23718 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
23719 (ira_sort_regnos_for_alter_reg): Likewise.
23720 * ira-costs.c (record_operand_costs): Likewise.
23721 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
23722 (resolve_simple_move): Likewise.
23723 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
23724 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
23725 (lra_constraints): Likewise.
23726 (CONST_POOL_OK_P): Reject variable-sized modes.
23727 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
23728 (add_pseudo_to_slot, lra_spill): Likewise.
23729 * omp-low.c (omp_clause_aligned_alignment): Likewise.
23730 * optabs-query.c (get_best_extraction_insn): Likewise.
23731 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
23732 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
23733 (expand_mult_highpart, valid_multiword_target_p): Likewise.
23734 * recog.c (offsettable_address_addr_space_p): Likewise.
23735 * regcprop.c (maybe_mode_change): Likewise.
23736 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
23737 * regrename.c (build_def_use): Likewise.
23738 * regstat.c (dump_reg_info): Likewise.
23739 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
23740 (find_reloads, find_reloads_subreg_address): Likewise.
23741 * reload1.c (eliminate_regs_1): Likewise.
23742 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
23743 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
23744 (simplify_binary_operation_1, simplify_subreg): Likewise.
23745 * targhooks.c (default_function_arg_padding): Likewise.
23746 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
23747 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
23748 (verify_gimple_assign_ternary): Likewise.
23749 * tree-inline.c (estimate_move_cost): Likewise.
23750 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23751 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
23752 (get_address_cost_ainc): Likewise.
23753 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
23754 (vect_supportable_dr_alignment): Likewise.
23755 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
23756 (vectorizable_reduction): Likewise.
23757 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
23758 (vectorizable_operation, vectorizable_load): Likewise.
23759 * tree.c (build_same_sized_truth_vector_type): Likewise.
23760 * valtrack.c (cleanup_auto_inc_dec): Likewise.
23761 * var-tracking.c (emit_note_insn_var_location): Likewise.
23762 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
23763 (ADDR_VEC_ALIGN): Likewise.
23765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23766 Alan Hayward <alan.hayward@arm.com>
23767 David Sherwood <david.sherwood@arm.com>
23769 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
23771 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
23772 or if measurement_type is polynomial.
23773 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
23774 * combine.c (make_extraction): Likewise.
23775 * dse.c (find_shift_sequence): Likewise.
23776 * dwarf2out.c (mem_loc_descriptor): Likewise.
23777 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
23778 (extract_bit_field, extract_low_bits): Likewise.
23779 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
23780 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
23781 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
23782 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
23783 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
23784 * reload.c (find_reloads): Likewise.
23785 * reload1.c (alter_reg): Likewise.
23786 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
23787 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
23788 * tree-if-conv.c (predicate_mem_writes): Likewise.
23789 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
23790 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
23791 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
23792 * valtrack.c (dead_debug_insert_temp): Likewise.
23793 * varasm.c (mergeable_constant_section): Likewise.
23794 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
23796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23797 Alan Hayward <alan.hayward@arm.com>
23798 David Sherwood <david.sherwood@arm.com>
23800 * expr.c (expand_assignment): Cope with polynomial mode sizes
23801 when assigning to a CONCAT.
23803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23804 Alan Hayward <alan.hayward@arm.com>
23805 David Sherwood <david.sherwood@arm.com>
23807 * machmode.h (mode_precision): Change from unsigned short to
23809 (mode_to_precision): Return a poly_uint16 rather than an unsigned
23811 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
23812 or if measurement_type is not polynomial.
23813 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
23814 in which the mode is already known to be a scalar_int_mode.
23815 * genmodes.c (emit_mode_precision): Change the type of mode_precision
23816 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
23818 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
23819 for GET_MODE_PRECISION.
23820 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
23821 for GET_MODE_PRECISION.
23822 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
23824 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
23825 (expand_field_assignment, make_extraction): Likewise.
23826 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
23827 (get_last_value): Likewise.
23828 * convert.c (convert_to_integer_1): Likewise.
23829 * cse.c (cse_insn): Likewise.
23830 * expr.c (expand_expr_real_1): Likewise.
23831 * lra-constraints.c (simplify_operand_subreg): Likewise.
23832 * optabs-query.c (can_atomic_load_p): Likewise.
23833 * optabs.c (expand_atomic_load): Likewise.
23834 (expand_atomic_store): Likewise.
23835 * ree.c (combine_reaching_defs): Likewise.
23836 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
23837 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
23838 * tree.h (type_has_mode_precision_p): Likewise.
23839 * ubsan.c (instrument_si_overflow): Likewise.
23841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23842 Alan Hayward <alan.hayward@arm.com>
23843 David Sherwood <david.sherwood@arm.com>
23845 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
23846 polynomial numbers of units.
23847 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
23848 (valid_vector_subparts_p): New function.
23849 (build_vector_type): Remove temporary shim and take the number
23850 of units as a poly_uint64 rather than an int.
23851 (build_opaque_vector_type): Take the number of units as a
23852 poly_uint64 rather than an int.
23853 * tree.c (build_vector_from_ctor): Handle polynomial
23854 TYPE_VECTOR_SUBPARTS.
23855 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
23856 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
23857 (build_vector_from_val): If the number of units is variable,
23858 use build_vec_duplicate_cst for constant operands and
23859 VEC_DUPLICATE_EXPR otherwise.
23860 (make_vector_type): Remove temporary is_constant ().
23861 (build_vector_type, build_opaque_vector_type): Take the number of
23862 units as a poly_uint64 rather than an int.
23863 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
23865 * cfgexpand.c (expand_debug_expr): Likewise.
23866 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
23867 (store_constructor, expand_expr_real_1): Likewise.
23868 (const_scalar_mask_from_tree): Likewise.
23869 * fold-const-call.c (fold_const_reduction): Likewise.
23870 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
23871 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
23872 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
23873 (fold_relational_const): Likewise.
23874 (native_interpret_vector): Likewise. Change the size from an
23875 int to an unsigned int.
23876 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
23877 TYPE_VECTOR_SUBPARTS.
23878 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
23879 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
23880 duplicating a non-constant operand into a variable-length vector.
23881 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
23882 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
23883 * ipa-icf.c (sem_variable::equals): Likewise.
23884 * match.pd: Likewise.
23885 * omp-simd-clone.c (simd_clone_subparts): Likewise.
23886 * print-tree.c (print_node): Likewise.
23887 * stor-layout.c (layout_type): Likewise.
23888 * targhooks.c (default_builtin_vectorization_cost): Likewise.
23889 * tree-cfg.c (verify_gimple_comparison): Likewise.
23890 (verify_gimple_assign_binary): Likewise.
23891 (verify_gimple_assign_ternary): Likewise.
23892 (verify_gimple_assign_single): Likewise.
23893 * tree-pretty-print.c (dump_generic_node): Likewise.
23894 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23895 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
23896 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
23897 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
23898 (vect_shift_permute_load_chain): Likewise.
23899 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
23900 (expand_vector_condition, optimize_vector_constructor): Likewise.
23901 (lower_vec_perm, get_compute_type): Likewise.
23902 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
23903 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
23904 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
23905 (vect_recog_mask_conversion_pattern): Likewise.
23906 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
23907 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
23908 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23909 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
23910 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
23911 (vectorizable_shift, vectorizable_operation, vectorizable_store)
23912 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
23913 (supportable_widening_operation): Likewise.
23914 (supportable_narrowing_operation): Likewise.
23915 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
23917 * varasm.c (output_constant): Likewise.
23919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23920 Alan Hayward <alan.hayward@arm.com>
23921 David Sherwood <david.sherwood@arm.com>
23923 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
23924 so that both the length == 3 and length != 3 cases set up their
23925 own permute vectors. Add comments explaining why we know the
23926 number of elements is constant.
23927 (vect_permute_load_chain): Likewise.
23929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23930 Alan Hayward <alan.hayward@arm.com>
23931 David Sherwood <david.sherwood@arm.com>
23933 * machmode.h (mode_nunits): Change from unsigned char to
23935 (ONLY_FIXED_SIZE_MODES): New macro.
23936 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
23937 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
23938 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
23940 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
23941 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
23942 or if measurement_type is not polynomial.
23943 * genmodes.c (ZERO_COEFFS): New macro.
23944 (emit_mode_nunits_inline): Make mode_nunits_inline return a
23946 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
23947 Use ZERO_COEFFS when emitting initializers.
23948 * data-streamer.h (bp_pack_poly_value): New function.
23949 (bp_unpack_poly_value): Likewise.
23950 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
23951 for GET_MODE_NUNITS.
23952 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
23953 for GET_MODE_NUNITS.
23954 * tree.c (make_vector_type): Remove temporary shim and make
23955 the real function take the number of units as a poly_uint64
23956 rather than an int.
23957 (build_vector_type_for_mode): Handle polynomial nunits.
23958 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
23959 * emit-rtl.c (const_vec_series_p_1): Likewise.
23960 (gen_rtx_CONST_VECTOR): Likewise.
23961 * fold-const.c (test_vec_duplicate_folding): Likewise.
23962 * genrecog.c (validate_pattern): Likewise.
23963 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
23964 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
23965 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
23966 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
23967 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
23968 * rtlanal.c (subreg_get_info): Likewise.
23969 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23970 (vect_grouped_load_supported): Likewise.
23971 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
23972 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
23973 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
23974 (simplify_const_unary_operation, simplify_binary_operation_1)
23975 (simplify_const_binary_operation, simplify_ternary_operation)
23976 (test_vector_ops_duplicate, test_vector_ops): Likewise.
23977 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
23978 instead of CONST_VECTOR_NUNITS.
23979 * varasm.c (output_constant_pool_2): Likewise.
23980 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
23981 explicit-encoded elements in the XVEC for variable-length vectors.
23983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23985 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
23987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23988 Alan Hayward <alan.hayward@arm.com>
23989 David Sherwood <david.sherwood@arm.com>
23991 * coretypes.h (fixed_size_mode): Declare.
23992 (fixed_size_mode_pod): New typedef.
23993 * builtins.h (target_builtins::x_apply_args_mode)
23994 (target_builtins::x_apply_result_mode): Change type to
23995 fixed_size_mode_pod.
23996 * builtins.c (apply_args_size, apply_result_size, result_vector)
23997 (expand_builtin_apply_args_1, expand_builtin_apply)
23998 (expand_builtin_return): Update accordingly.
24000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24002 * cse.c (hash_rtx_cb): Hash only the encoded elements.
24003 * cselib.c (cselib_hash_rtx): Likewise.
24004 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
24005 CONST_VECTOR encoding.
24007 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24008 Jeff Law <law@redhat.com>
24011 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
24012 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
24013 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
24014 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
24017 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
24018 explicitly probe *sp in a noreturn function if there were any callee
24019 register saves or frame pointer is needed.
24021 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24024 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
24025 BLKmode for ternary, binary or unary expressions.
24028 * var-tracking.c (delete_vta_debug_insn): New inline function.
24029 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
24030 insns from get_insns () to NULL instead of each bb separately.
24031 Use delete_vta_debug_insn. No longer static.
24032 (vt_debug_insns_local, variable_tracking_main_1): Adjust
24033 delete_vta_debug_insns callers.
24034 * rtl.h (delete_vta_debug_insns): Declare.
24035 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
24036 instead of variable_tracking_main.
24038 2018-01-03 Martin Sebor <msebor@redhat.com>
24040 PR tree-optimization/83603
24041 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
24042 arguments past the endof the argument list in functions declared
24043 without a prototype.
24044 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
24045 Avoid checking when arguments are null.
24047 2018-01-03 Martin Sebor <msebor@redhat.com>
24050 * doc/extend.texi (attribute const): Fix a typo.
24051 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
24052 issuing -Wsuggest-attribute for void functions.
24054 2018-01-03 Martin Sebor <msebor@redhat.com>
24056 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
24057 offset_int::from instead of wide_int::to_shwi.
24058 (maybe_diag_overlap): Remove assertion.
24059 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
24060 * gimple-ssa-sprintf.c (format_directive): Same.
24061 (parse_directive): Same.
24062 (sprintf_dom_walker::compute_format_length): Same.
24063 (try_substitute_return_value): Same.
24065 2018-01-03 Jeff Law <law@redhat.com>
24067 PR middle-end/83654
24068 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
24069 non-constant residual for zero at runtime and avoid probing in
24070 that case. Reorganize code for trailing problem to mirror handling
24073 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24075 PR tree-optimization/83501
24076 * tree-ssa-strlen.c (get_string_cst): New.
24077 (handle_char_store): Call get_string_cst.
24079 2018-01-03 Martin Liska <mliska@suse.cz>
24081 PR tree-optimization/83593
24082 * tree-ssa-strlen.c: Include tree-cfg.h.
24083 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
24084 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
24085 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
24087 (strlen_dom_walker::before_dom_children): Call
24088 gimple_purge_dead_eh_edges. Dump tranformation with details
24090 (strlen_dom_walker::before_dom_children): Update call by adding
24091 new argument cleanup_eh.
24092 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
24094 2018-01-03 Martin Liska <mliska@suse.cz>
24097 * cif-code.def (VARIADIC_THUNK): New enum value.
24098 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
24101 2018-01-03 Jan Beulich <jbeulich@suse.com>
24103 * sse.md (mov<mode>_internal): Tighten condition for when to use
24104 vmovdqu<ssescalarsize> for TI and OI modes.
24106 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24108 Update copyright years.
24110 2018-01-03 Martin Liska <mliska@suse.cz>
24113 * ipa-visibility.c (function_and_variable_visibility): Skip
24114 functions with noipa attribure.
24116 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24118 * gcc.c (process_command): Update copyright notice dates.
24119 * gcov-dump.c (print_version): Ditto.
24120 * gcov.c (print_version): Ditto.
24121 * gcov-tool.c (print_version): Ditto.
24122 * gengtype.c (create_file): Ditto.
24123 * doc/cpp.texi: Bump @copying's copyright year.
24124 * doc/cppinternals.texi: Ditto.
24125 * doc/gcc.texi: Ditto.
24126 * doc/gccint.texi: Ditto.
24127 * doc/gcov.texi: Ditto.
24128 * doc/install.texi: Ditto.
24129 * doc/invoke.texi: Ditto.
24131 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24133 * vector-builder.h (vector_builder::m_full_nelts): Change from
24134 unsigned int to poly_uint64.
24135 (vector_builder::full_nelts): Update prototype accordingly.
24136 (vector_builder::new_vector): Likewise.
24137 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
24138 (vector_builder::operator ==): Likewise.
24139 (vector_builder::finalize): Likewise.
24140 * int-vector-builder.h (int_vector_builder::int_vector_builder):
24141 Take the number of elements as a poly_uint64 rather than an
24143 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
24144 from unsigned int to poly_uint64.
24145 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
24146 (vec_perm_indices::new_vector): Likewise.
24147 (vec_perm_indices::length): Likewise.
24148 (vec_perm_indices::nelts_per_input): Likewise.
24149 (vec_perm_indices::input_nelts): Likewise.
24150 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
24151 number of elements per input as a poly_uint64 rather than an
24152 unsigned int. Use the original encoding for variable-length
24153 vectors, rather than clamping each individual element.
24154 For the second and subsequent elements in each pattern,
24155 clamp the step and base before clamping their sum.
24156 (vec_perm_indices::series_p): Handle polynomial element counts.
24157 (vec_perm_indices::all_in_range_p): Likewise.
24158 (vec_perm_indices_to_tree): Likewise.
24159 (vec_perm_indices_to_rtx): Likewise.
24160 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
24161 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
24162 (tree_vector_builder::new_binary_operation): Handle polynomial
24163 element counts. Return false if we need to know the number
24164 of elements at compile time.
24165 * fold-const.c (fold_vec_perm): Punt if the number of elements
24166 isn't known at compile time.
24168 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24170 * vec-perm-indices.h (vec_perm_builder): Change element type
24171 from HOST_WIDE_INT to poly_int64.
24172 (vec_perm_indices::element_type): Update accordingly.
24173 (vec_perm_indices::clamp): Handle polynomial element_types.
24174 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
24175 (vec_perm_indices::all_in_range_p): Likewise.
24176 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
24178 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
24179 polynomial vec_perm_indices element types.
24180 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
24181 * fold-const.c (fold_vec_perm): Likewise.
24182 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
24183 * tree-vect-generic.c (lower_vec_perm): Likewise.
24184 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
24185 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
24186 element type to HOST_WIDE_INT.
24188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24189 Alan Hayward <alan.hayward@arm.com>
24190 David Sherwood <david.sherwood@arm.com>
24192 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
24193 rather than an int. Use plus_constant.
24194 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
24195 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
24197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24198 Alan Hayward <alan.hayward@arm.com>
24199 David Sherwood <david.sherwood@arm.com>
24201 * calls.c (emit_call_1, expand_call): Change struct_value_size from
24202 a HOST_WIDE_INT to a poly_int64.
24204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24205 Alan Hayward <alan.hayward@arm.com>
24206 David Sherwood <david.sherwood@arm.com>
24208 * calls.c (load_register_parameters): Cope with polynomial
24209 mode sizes. Require a constant size for BLKmode parameters
24210 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
24211 forces a parameter to be padded at the lsb end in order to
24212 fill a complete number of words, require the parameter size
24213 to be ordered wrt UNITS_PER_WORD.
24215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24216 Alan Hayward <alan.hayward@arm.com>
24217 David Sherwood <david.sherwood@arm.com>
24219 * reload1.c (spill_stack_slot_width): Change element type
24220 from unsigned int to poly_uint64_pod.
24221 (alter_reg): Treat mode sizes as polynomial.
24223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24224 Alan Hayward <alan.hayward@arm.com>
24225 David Sherwood <david.sherwood@arm.com>
24227 * reload.c (complex_word_subreg_p): New function.
24228 (reload_inner_reg_of_subreg, push_reload): Use it.
24230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24231 Alan Hayward <alan.hayward@arm.com>
24232 David Sherwood <david.sherwood@arm.com>
24234 * lra-constraints.c (process_alt_operands): Reject matched
24235 operands whose sizes aren't ordered.
24236 (match_reload): Refer to this check here.
24238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24239 Alan Hayward <alan.hayward@arm.com>
24240 David Sherwood <david.sherwood@arm.com>
24242 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
24243 that the mode size is in the set {1, 2, 4, 8, 16}.
24245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24246 Alan Hayward <alan.hayward@arm.com>
24247 David Sherwood <david.sherwood@arm.com>
24249 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
24250 Use plus_constant instead of gen_rtx_PLUS.
24252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24253 Alan Hayward <alan.hayward@arm.com>
24254 David Sherwood <david.sherwood@arm.com>
24256 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
24257 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
24258 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
24259 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
24260 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
24261 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
24262 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
24263 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
24264 * config/i386/i386.c (ix86_push_rounding): ...this new function.
24265 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
24267 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
24268 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
24269 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
24270 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
24271 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
24272 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
24273 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
24274 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
24275 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
24276 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
24278 * expr.c (emit_move_resolve_push): Treat the input and result
24279 of PUSH_ROUNDING as a poly_int64.
24280 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
24281 (emit_push_insn): Likewise.
24282 * lra-eliminations.c (mark_not_eliminable): Likewise.
24283 * recog.c (push_operand): Likewise.
24284 * reload1.c (elimination_effects): Likewise.
24285 * rtlanal.c (nonzero_bits1): Likewise.
24286 * calls.c (store_one_arg): Likewise. Require the padding to be
24287 known at compile time.
24289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24290 Alan Hayward <alan.hayward@arm.com>
24291 David Sherwood <david.sherwood@arm.com>
24293 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
24294 Use plus_constant instead of gen_rtx_PLUS.
24296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24297 Alan Hayward <alan.hayward@arm.com>
24298 David Sherwood <david.sherwood@arm.com>
24300 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
24301 rather than an int.
24303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24304 Alan Hayward <alan.hayward@arm.com>
24305 David Sherwood <david.sherwood@arm.com>
24307 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
24308 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
24309 via stack temporaries. Treat the mode size as polynomial too.
24311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24312 Alan Hayward <alan.hayward@arm.com>
24313 David Sherwood <david.sherwood@arm.com>
24315 * expr.c (expand_expr_real_2): When handling conversions involving
24316 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
24317 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
24318 as a poly_uint64 too.
24320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24321 Alan Hayward <alan.hayward@arm.com>
24322 David Sherwood <david.sherwood@arm.com>
24324 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
24326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24327 Alan Hayward <alan.hayward@arm.com>
24328 David Sherwood <david.sherwood@arm.com>
24330 * combine.c (can_change_dest_mode): Handle polynomial
24331 REGMODE_NATURAL_SIZE.
24332 * expmed.c (store_bit_field_1): Likewise.
24333 * expr.c (store_constructor): Likewise.
24334 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
24335 and polynomial REGMODE_NATURAL_SIZE.
24336 (gen_lowpart_common): Likewise.
24337 * reginfo.c (record_subregs_of_mode): Likewise.
24338 * rtlanal.c (read_modify_subreg_p): Likewise.
24340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24341 Alan Hayward <alan.hayward@arm.com>
24342 David Sherwood <david.sherwood@arm.com>
24344 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
24345 numbers of elements.
24347 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24348 Alan Hayward <alan.hayward@arm.com>
24349 David Sherwood <david.sherwood@arm.com>
24351 * match.pd: Cope with polynomial numbers of vector elements.
24353 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24354 Alan Hayward <alan.hayward@arm.com>
24355 David Sherwood <david.sherwood@arm.com>
24357 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
24358 in a POINTER_PLUS_EXPR.
24360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24361 Alan Hayward <alan.hayward@arm.com>
24362 David Sherwood <david.sherwood@arm.com>
24364 * omp-simd-clone.c (simd_clone_subparts): New function.
24365 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
24366 (ipa_simd_modify_function_body): Likewise.
24368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24369 Alan Hayward <alan.hayward@arm.com>
24370 David Sherwood <david.sherwood@arm.com>
24372 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
24373 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
24374 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
24375 (expand_vector_condition, vector_element): Likewise.
24376 (subparts_gt): New function.
24377 (get_compute_type): Use subparts_gt.
24378 (count_type_subparts): Delete.
24379 (expand_vector_operations_1): Use subparts_gt instead of
24380 count_type_subparts.
24382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24383 Alan Hayward <alan.hayward@arm.com>
24384 David Sherwood <david.sherwood@arm.com>
24386 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
24387 (vect_compile_time_alias): ...this new function. Do the calculation
24388 on poly_ints rather than trees.
24389 (vect_prune_runtime_alias_test_list): Update call accordingly.
24391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24392 Alan Hayward <alan.hayward@arm.com>
24393 David Sherwood <david.sherwood@arm.com>
24395 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
24397 (vect_schedule_slp_instance): Likewise.
24399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24400 Alan Hayward <alan.hayward@arm.com>
24401 David Sherwood <david.sherwood@arm.com>
24403 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
24404 constant and extern definitions for variable-length vectors.
24405 (vect_get_constant_vectors): Note that the number of units
24406 is known to be constant.
24408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24409 Alan Hayward <alan.hayward@arm.com>
24410 David Sherwood <david.sherwood@arm.com>
24412 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
24413 of units as polynomial. Choose between WIDE and NARROW based
24416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24417 Alan Hayward <alan.hayward@arm.com>
24418 David Sherwood <david.sherwood@arm.com>
24420 * tree-vect-stmts.c (simd_clone_subparts): New function.
24421 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
24423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24424 Alan Hayward <alan.hayward@arm.com>
24425 David Sherwood <david.sherwood@arm.com>
24427 * tree-vect-stmts.c (vectorizable_call): Treat the number of
24428 vectors as polynomial. Use build_index_vector for
24429 IFN_GOMP_SIMD_LANE.
24431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24432 Alan Hayward <alan.hayward@arm.com>
24433 David Sherwood <david.sherwood@arm.com>
24435 * tree-vect-stmts.c (get_load_store_type): Treat the number of
24436 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
24437 for variable-length vectors.
24438 (vectorizable_mask_load_store): Treat the number of units as
24439 polynomial, asserting that it is constant if the condition has
24440 already been enforced.
24441 (vectorizable_store, vectorizable_load): Likewise.
24443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24444 Alan Hayward <alan.hayward@arm.com>
24445 David Sherwood <david.sherwood@arm.com>
24447 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
24448 of units as polynomial. Punt if we can't tell at compile time
24449 which vector contains the final result.
24451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24452 Alan Hayward <alan.hayward@arm.com>
24453 David Sherwood <david.sherwood@arm.com>
24455 * tree-vect-loop.c (vectorizable_induction): Treat the number
24456 of units as polynomial. Punt on SLP inductions. Use an integer
24457 VEC_SERIES_EXPR for variable-length integer reductions. Use a
24458 cast of such a series for variable-length floating-point
24461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24462 Alan Hayward <alan.hayward@arm.com>
24463 David Sherwood <david.sherwood@arm.com>
24465 * tree.h (build_index_vector): Declare.
24466 * tree.c (build_index_vector): New function.
24467 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
24468 of units as polynomial, forcibly converting it to a constant if
24469 vectorizable_reduction has already enforced the condition.
24470 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
24471 to create a {1,2,3,...} vector.
24472 (vectorizable_reduction): Treat the number of units as polynomial.
24473 Choose vectype_in based on the largest scalar element size rather
24474 than the smallest number of units. Enforce the restrictions
24477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24478 Alan Hayward <alan.hayward@arm.com>
24479 David Sherwood <david.sherwood@arm.com>
24481 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
24482 number of units as polynomial.
24484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24485 Alan Hayward <alan.hayward@arm.com>
24486 David Sherwood <david.sherwood@arm.com>
24488 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
24489 * target.def (autovectorize_vector_sizes): Return the vector sizes
24490 by pointer, using vector_sizes rather than a bitmask.
24491 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
24492 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
24493 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
24495 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
24496 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
24497 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
24498 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
24499 * omp-general.c (omp_max_vf): Likewise.
24500 * omp-low.c (omp_clause_aligned_alignment): Likewise.
24501 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
24502 * tree-vect-loop.c (vect_analyze_loop): Likewise.
24503 * tree-vect-slp.c (vect_slp_bb): Likewise.
24504 * doc/tm.texi: Regenerate.
24505 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
24507 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
24508 the vector size as a poly_uint64 rather than an unsigned int.
24509 (current_vector_size): Change from an unsigned int to a poly_uint64.
24510 (get_vectype_for_scalar_type): Update accordingly.
24511 * tree.h (build_truth_vector_type): Take the size and number of
24512 units as a poly_uint64 rather than an unsigned int.
24513 (build_vector_type): Add a temporary overload that takes
24514 the number of units as a poly_uint64 rather than an unsigned int.
24515 * tree.c (make_vector_type): Likewise.
24516 (build_truth_vector_type): Take the number of units as a poly_uint64
24517 rather than an unsigned int.
24519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24520 Alan Hayward <alan.hayward@arm.com>
24521 David Sherwood <david.sherwood@arm.com>
24523 * target.def (get_mask_mode): Take the number of units and length
24524 as poly_uint64s rather than unsigned ints.
24525 * targhooks.h (default_get_mask_mode): Update accordingly.
24526 * targhooks.c (default_get_mask_mode): Likewise.
24527 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
24528 * doc/tm.texi: Regenerate.
24530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24531 Alan Hayward <alan.hayward@arm.com>
24532 David Sherwood <david.sherwood@arm.com>
24534 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
24535 * omp-general.c (omp_max_vf): Likewise.
24536 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
24537 (expand_omp_simd): Handle polynomial safelen.
24538 * omp-low.c (omplow_simd_context): Add a default constructor.
24539 (omplow_simd_context::max_vf): Change from int to poly_uint64.
24540 (lower_rec_simd_input_clauses): Update accordingly.
24541 (lower_rec_input_clauses): Likewise.
24543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24544 Alan Hayward <alan.hayward@arm.com>
24545 David Sherwood <david.sherwood@arm.com>
24547 * tree-vectorizer.h (vect_nunits_for_cost): New function.
24548 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
24549 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
24550 (vect_analyze_slp_cost): Likewise.
24551 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
24552 (vect_model_load_cost): Likewise.
24554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24555 Alan Hayward <alan.hayward@arm.com>
24556 David Sherwood <david.sherwood@arm.com>
24558 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
24559 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
24560 from an unsigned int * to a poly_uint64_pod *.
24561 (calculate_unrolling_factor): New function.
24562 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
24564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24565 Alan Hayward <alan.hayward@arm.com>
24566 David Sherwood <david.sherwood@arm.com>
24568 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
24569 from an unsigned int to a poly_uint64.
24570 (_loop_vec_info::slp_unrolling_factor): Likewise.
24571 (_loop_vec_info::vectorization_factor): Change from an int
24573 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
24574 (vect_get_num_vectors): New function.
24575 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
24576 (vect_get_num_copies): Use vect_get_num_vectors.
24577 (vect_analyze_data_ref_dependences): Change max_vf from an int *
24578 to an unsigned int *.
24579 (vect_analyze_data_refs): Change min_vf from an int * to a
24581 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
24582 than an unsigned HOST_WIDE_INT.
24583 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
24584 (vect_analyze_data_ref_dependence): Change max_vf from an int *
24585 to an unsigned int *.
24586 (vect_analyze_data_ref_dependences): Likewise.
24587 (vect_compute_data_ref_alignment): Handle polynomial vf.
24588 (vect_enhance_data_refs_alignment): Likewise.
24589 (vect_prune_runtime_alias_test_list): Likewise.
24590 (vect_shift_permute_load_chain): Likewise.
24591 (vect_supportable_dr_alignment): Likewise.
24592 (dependence_distance_ge_vf): Take the vectorization factor as a
24593 poly_uint64 rather than an unsigned HOST_WIDE_INT.
24594 (vect_analyze_data_refs): Change min_vf from an int * to a
24596 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
24597 vfm1 as a poly_uint64 rather than an int. Make the same change
24598 for the returned bound_scalar.
24599 (vect_gen_vector_loop_niters): Handle polynomial vf.
24600 (vect_do_peeling): Likewise. Update call to
24601 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
24602 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
24604 * tree-vect-loop.c (vect_determine_vectorization_factor)
24605 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
24606 (vect_get_known_peeling_cost): Likewise.
24607 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
24608 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
24609 (vect_transform_loop): Likewise. Use the lowest possible VF when
24610 updating the upper bounds of the loop.
24611 (vect_min_worthwhile_factor): Make static. Return an unsigned int
24612 rather than an int.
24613 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
24614 polynomial unroll factors.
24615 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
24616 (vect_make_slp_decision): Likewise.
24617 (vect_supported_load_permutation_p): Likewise, and polynomial
24619 (vect_analyze_slp_cost): Handle polynomial vf.
24620 (vect_slp_analyze_node_operations): Likewise.
24621 (vect_slp_analyze_bb_1): Likewise.
24622 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
24623 than an unsigned HOST_WIDE_INT.
24624 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
24625 (vectorizable_load): Handle polynomial vf.
24626 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
24628 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
24630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24631 Alan Hayward <alan.hayward@arm.com>
24632 David Sherwood <david.sherwood@arm.com>
24634 * match.pd: Handle bit operations involving three constants
24635 and try to fold one pair.
24637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24639 * tree-vect-loop-manip.c: Include gimple-fold.h.
24640 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
24641 niters_maybe_zero parameters. Handle other cases besides a step of 1.
24642 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
24643 Add a path that uses a step of VF instead of 1, but disable it
24645 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
24646 and niters_no_overflow parameters. Update calls to
24647 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
24648 Create a new SSA name if the latter choses to use a ste other
24649 than zero, and return it via niters_vector_mult_vf_var.
24650 * tree-vect-loop.c (vect_transform_loop): Update calls to
24651 vect_do_peeling, vect_gen_vector_loop_niters and
24652 slpeel_make_loop_iterate_ntimes.
24653 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
24654 (vect_gen_vector_loop_niters): Update declarations after above changes.
24656 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
24658 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
24659 128-bit round to integer instructions.
24660 (ceil<mode>2): Likewise.
24661 (btrunc<mode>2): Likewise.
24662 (round<mode>2): Likewise.
24664 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24666 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
24667 unaligned VSX load/store on P8/P9.
24668 (expand_block_clear): Allow the use of unaligned VSX
24669 load/store on P8/P9.
24671 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24673 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
24675 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
24676 swap associated with both a load and a store.
24678 2018-01-02 Andrew Waterman <andrew@sifive.com>
24680 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
24681 * config/riscv/riscv.md (clear_cache): Use it.
24683 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
24685 * web.c: Remove out-of-date comment.
24687 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24689 * expr.c (fixup_args_size_notes): Check that any existing
24690 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
24691 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
24692 (emit_single_push_insn): ...here.
24694 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24696 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
24697 (const_vector_encoded_nelts): New function.
24698 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
24699 (const_vector_int_elt, const_vector_elt): Declare.
24700 * emit-rtl.c (const_vector_int_elt_1): New function.
24701 (const_vector_elt): Likewise.
24702 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
24703 of CONST_VECTOR_ELT.
24705 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24707 * expr.c: Include rtx-vector-builder.h.
24708 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
24709 directly on the tree encoding.
24710 (const_vector_from_tree): Likewise.
24711 * optabs.c: Include rtx-vector-builder.h.
24712 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
24713 sequence of "u" values.
24714 * vec-perm-indices.c: Include rtx-vector-builder.h.
24715 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
24716 directly on the vec_perm_indices encoding.
24718 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24720 * doc/rtl.texi (const_vector): Describe new encoding scheme.
24721 * Makefile.in (OBJS): Add rtx-vector-builder.o.
24722 * rtx-vector-builder.h: New file.
24723 * rtx-vector-builder.c: Likewise.
24724 * rtl.h (rtx_def::u2): Add a const_vector field.
24725 (CONST_VECTOR_NPATTERNS): New macro.
24726 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
24727 (CONST_VECTOR_DUPLICATE_P): Likewise.
24728 (CONST_VECTOR_STEPPED_P): Likewise.
24729 (CONST_VECTOR_ENCODED_ELT): Likewise.
24730 (const_vec_duplicate_p): Check for a duplicated vector encoding.
24731 (unwrap_const_vec_duplicate): Likewise.
24732 (const_vec_series_p): Check for a non-duplicated vector encoding.
24733 Say that the function only returns true for integer vectors.
24734 * emit-rtl.c: Include rtx-vector-builder.h.
24735 (gen_const_vec_duplicate_1): Delete.
24736 (gen_const_vector): Call gen_const_vec_duplicate instead of
24737 gen_const_vec_duplicate_1.
24738 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
24739 (gen_const_vec_duplicate): Use rtx_vector_builder.
24740 (gen_const_vec_series): Likewise.
24741 (gen_rtx_CONST_VECTOR): Likewise.
24742 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
24743 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
24744 Build a new vector rather than modifying a CONST_VECTOR in-place.
24745 (handle_special_swappables): Update call accordingly.
24746 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
24747 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
24748 Build a new vector rather than modifying a CONST_VECTOR in-place.
24749 (handle_special_swappables): Update call accordingly.
24751 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24753 * simplify-rtx.c (simplify_const_binary_operation): Use
24754 CONST_VECTOR_ELT instead of XVECEXP.
24756 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24758 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
24759 the selector elements to be different from the data elements
24760 if the selector is a VECTOR_CST.
24761 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
24762 ssizetype for the selector.
24764 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24766 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
24767 before testing each element individually.
24768 * tree-vect-generic.c (lower_vec_perm): Likewise.
24770 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24772 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
24773 * selftest-run-tests.c (selftest::run_tests): Call it.
24774 * vector-builder.h (vector_builder::operator ==): New function.
24775 (vector_builder::operator !=): Likewise.
24776 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
24777 (vec_perm_indices::all_from_input_p): New function.
24778 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
24779 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
24780 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
24781 instead of reading the VECTOR_CST directly. Detect whether both
24782 vector inputs are the same before constructing the vec_perm_indices,
24783 and update the number of inputs argument accordingly. Use the
24784 utility functions added above. Only construct sel2 if we need to.
24786 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24788 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
24789 the broadcast of the low byte.
24790 (expand_mult_highpart): Use an explicit encoding for the permutes.
24791 * optabs-query.c (can_mult_highpart_p): Likewise.
24792 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
24793 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24794 (vectorizable_bswap): Likewise.
24795 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
24796 explicit encoding for the power-of-2 permutes.
24797 (vect_permute_store_chain): Likewise.
24798 (vect_grouped_load_supported): Likewise.
24799 (vect_permute_load_chain): Likewise.
24801 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24803 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
24804 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
24805 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
24806 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
24807 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
24808 (vect_gen_perm_mask_any): Likewise.
24810 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24812 * int-vector-builder.h: New file.
24813 * vec-perm-indices.h: Include int-vector-builder.h.
24814 (vec_perm_indices): Redefine as an int_vector_builder.
24815 (auto_vec_perm_indices): Delete.
24816 (vec_perm_builder): Redefine as a stand-alone class.
24817 (vec_perm_indices::vec_perm_indices): New function.
24818 (vec_perm_indices::clamp): Likewise.
24819 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
24820 (vec_perm_indices::new_vector): New function.
24821 (vec_perm_indices::new_expanded_vector): Update for new
24822 vec_perm_indices class.
24823 (vec_perm_indices::rotate_inputs): New function.
24824 (vec_perm_indices::all_in_range_p): Operate directly on the
24825 encoded form, without computing elided elements.
24826 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
24827 encoding. Update for new vec_perm_indices class.
24828 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
24829 the given vec_perm_builder.
24830 (expand_vec_perm_var): Update vec_perm_builder constructor.
24831 (expand_mult_highpart): Use vec_perm_builder instead of
24832 auto_vec_perm_indices.
24833 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
24834 vec_perm_indices instead of auto_vec_perm_indices. Use a single
24835 or double series encoding as appropriate.
24836 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
24837 vec_perm_indices instead of auto_vec_perm_indices.
24838 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24839 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24840 (vect_permute_store_chain): Likewise.
24841 (vect_grouped_load_supported): Likewise.
24842 (vect_permute_load_chain): Likewise.
24843 (vect_shift_permute_load_chain): Likewise.
24844 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
24845 (vect_transform_slp_perm_load): Likewise.
24846 (vect_schedule_slp_instance): Likewise.
24847 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24848 (vectorizable_mask_load_store): Likewise.
24849 (vectorizable_bswap): Likewise.
24850 (vectorizable_store): Likewise.
24851 (vectorizable_load): Likewise.
24852 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
24853 vec_perm_indices instead of auto_vec_perm_indices. Use
24854 tree_to_vec_perm_builder to read the vector from a tree.
24855 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
24856 vec_perm_builder instead of a vec_perm_indices.
24857 (have_whole_vector_shift): Use vec_perm_builder and
24858 vec_perm_indices instead of auto_vec_perm_indices. Leave the
24859 truncation to calc_vec_perm_mask_for_shift.
24860 (vect_create_epilog_for_reduction): Likewise.
24861 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
24862 from auto_vec_perm_indices to vec_perm_indices.
24863 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
24864 instead of changing individual elements.
24865 (aarch64_vectorize_vec_perm_const): Use new_vector to install
24866 the vector in d.perm.
24867 * config/arm/arm.c (expand_vec_perm_d::perm): Change
24868 from auto_vec_perm_indices to vec_perm_indices.
24869 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
24870 instead of changing individual elements.
24871 (arm_vectorize_vec_perm_const): Use new_vector to install
24872 the vector in d.perm.
24873 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
24874 Update vec_perm_builder constructor.
24875 (rs6000_expand_interleave): Likewise.
24876 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
24877 (rs6000_expand_interleave): Likewise.
24879 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24881 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
24882 to qimode could truncate the indices.
24883 * optabs.c (expand_vec_perm_var): Likewise.
24885 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24887 * Makefile.in (OBJS): Add vec-perm-indices.o.
24888 * vec-perm-indices.h: New file.
24889 * vec-perm-indices.c: Likewise.
24890 * target.h (vec_perm_indices): Replace with a forward class
24892 (auto_vec_perm_indices): Move to vec-perm-indices.h.
24893 * optabs.h: Include vec-perm-indices.h.
24894 (expand_vec_perm): Delete.
24895 (selector_fits_mode_p, expand_vec_perm_var): Declare.
24896 (expand_vec_perm_const): Declare.
24897 * target.def (vec_perm_const_ok): Replace with...
24898 (vec_perm_const): ...this new hook.
24899 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
24900 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
24901 * doc/tm.texi: Regenerate.
24902 * optabs.def (vec_perm_const): Delete.
24903 * doc/md.texi (vec_perm_const): Likewise.
24904 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
24905 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
24906 expand_vec_perm for constant permutation vectors. Assert that
24907 the mode of variable permutation vectors is the integer equivalent
24908 of the mode that is being permuted.
24909 * optabs-query.h (selector_fits_mode_p): Declare.
24910 * optabs-query.c: Include vec-perm-indices.h.
24911 (selector_fits_mode_p): New function.
24912 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
24913 is defined, instead of checking whether the vec_perm_const_optab
24914 exists. Use targetm.vectorize.vec_perm_const instead of
24915 targetm.vectorize.vec_perm_const_ok. Check whether the indices
24916 fit in the vector mode before using a variable permute.
24917 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
24918 vec_perm_indices instead of an rtx.
24919 (expand_vec_perm): Replace with...
24920 (expand_vec_perm_const): ...this new function. Take the selector
24921 as a vec_perm_indices rather than an rtx. Also take the mode of
24922 the selector. Update call to shift_amt_for_vec_perm_mask.
24923 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
24924 Use vec_perm_indices::new_expanded_vector to expand the original
24925 selector into bytes. Check whether the indices fit in the vector
24926 mode before using a variable permute.
24927 (expand_vec_perm_var): Make global.
24928 (expand_mult_highpart): Use expand_vec_perm_const.
24929 * fold-const.c: Includes vec-perm-indices.h.
24930 * tree-ssa-forwprop.c: Likewise.
24931 * tree-vect-data-refs.c: Likewise.
24932 * tree-vect-generic.c: Likewise.
24933 * tree-vect-loop.c: Likewise.
24934 * tree-vect-slp.c: Likewise.
24935 * tree-vect-stmts.c: Likewise.
24936 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
24938 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
24939 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
24940 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
24941 (aarch64_vectorize_vec_perm_const): ...this new function.
24942 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
24943 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24944 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
24945 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
24946 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
24947 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24948 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
24950 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
24951 check for NEON modes.
24952 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
24953 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
24954 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
24955 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
24957 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
24958 the old VEC_PERM_CONST conditions.
24959 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
24960 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
24961 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
24962 (ia64_vectorize_vec_perm_const_ok): Merge into...
24963 (ia64_vectorize_vec_perm_const): ...this new function.
24964 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
24965 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
24966 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
24967 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
24968 * config/mips/mips.c (mips_expand_vec_perm_const)
24969 (mips_vectorize_vec_perm_const_ok): Merge into...
24970 (mips_vectorize_vec_perm_const): ...this new function.
24971 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
24972 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
24973 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
24974 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
24975 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
24976 (rs6000_expand_vec_perm_const): Delete.
24977 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
24979 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24980 (altivec_expand_vec_perm_const_le): Take each operand individually.
24981 Operate on constant selectors rather than rtxes.
24982 (altivec_expand_vec_perm_const): Likewise. Update call to
24983 altivec_expand_vec_perm_const_le.
24984 (rs6000_expand_vec_perm_const): Delete.
24985 (rs6000_vectorize_vec_perm_const_ok): Delete.
24986 (rs6000_vectorize_vec_perm_const): New function.
24987 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
24988 an element count and rtx array.
24989 (rs6000_expand_extract_even): Update call accordingly.
24990 (rs6000_expand_interleave): Likewise.
24991 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
24992 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
24993 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
24994 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
24995 (rs6000_expand_vec_perm_const): Delete.
24996 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
24997 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24998 (altivec_expand_vec_perm_const_le): Take each operand individually.
24999 Operate on constant selectors rather than rtxes.
25000 (altivec_expand_vec_perm_const): Likewise. Update call to
25001 altivec_expand_vec_perm_const_le.
25002 (rs6000_expand_vec_perm_const): Delete.
25003 (rs6000_vectorize_vec_perm_const_ok): Delete.
25004 (rs6000_vectorize_vec_perm_const): New function. Remove stray
25005 reference to the SPE evmerge intructions.
25006 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
25007 an element count and rtx array.
25008 (rs6000_expand_extract_even): Update call accordingly.
25009 (rs6000_expand_interleave): Likewise.
25010 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
25011 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
25013 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25015 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25017 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
25018 vector mode and that that mode matches the mode of the data
25020 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
25021 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
25022 directly using expand_vec_perm_1 when forcing selectors into
25024 (expand_vec_perm_var): New function, split out from expand_vec_perm.
25026 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25028 * optabs-query.h (can_vec_perm_p): Delete.
25029 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
25030 * optabs-query.c (can_vec_perm_p): Split into...
25031 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
25032 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
25033 particular selector is valid.
25034 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25035 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25036 (vect_grouped_load_supported): Likewise.
25037 (vect_shift_permute_load_chain): Likewise.
25038 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
25039 (vect_transform_slp_perm_load): Likewise.
25040 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25041 (vectorizable_bswap): Likewise.
25042 (vect_gen_perm_mask_checked): Likewise.
25043 * fold-const.c (fold_ternary_loc): Likewise. Don't take
25044 implementations of variable permutation vectors into account
25045 when deciding which selector to use.
25046 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
25047 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
25048 with a false third argument.
25049 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
25050 to test whether the constant selector is valid and can_vec_perm_var_p
25051 to test whether a variable selector is valid.
25053 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25055 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
25056 * optabs-query.c (can_vec_perm_p): Likewise.
25057 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
25058 instead of vec_perm_indices.
25059 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
25060 (vect_gen_perm_mask_checked): Likewise,
25061 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
25062 (vect_gen_perm_mask_checked): Likewise,
25064 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25066 * optabs-query.h (qimode_for_vec_perm): Declare.
25067 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
25068 (qimode_for_vec_perm): ...this new function.
25069 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
25071 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25073 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
25074 does not have a conditional at the top.
25076 2018-01-02 Richard Biener <rguenther@suse.de>
25078 * ipa-inline.c (big_speedup_p): Fix expression.
25080 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
25083 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
25086 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
25090 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
25091 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
25092 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
25093 cond_taken_branch_cost 3->4.
25095 2018-01-01 Jakub Jelinek <jakub@redhat.com>
25097 PR tree-optimization/83581
25098 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
25099 TODO_cleanup_cfg if any changes have been made.
25101 PR middle-end/83608
25102 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
25103 convert_modes if target mode has the right side, but different mode
25106 PR middle-end/83609
25107 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
25108 last argument when extracting from CONCAT. If either from_real or
25109 from_imag is NULL, use expansion through memory. If result is not
25110 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
25111 the parts directly to inner mode, if even that fails, use expansion
25114 PR middle-end/83623
25115 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
25116 check for bswap in mode rather than HImode and use that in expand_unop
25119 Copyright (C) 2018 Free Software Foundation, Inc.
25121 Copying and distribution of this file, with or without modification,
25122 are permitted in any medium without royalty provided the copyright
25123 notice and this notice are preserved.